- Use parameter packs for multicore mechanisms.
- Adjust unit tests and modcc.
- CPU and GPU use the same mechanism_ppack_base.
- CPU code uses a split similar to GPU with `mech_XYZ_foo` and `kernel::mech_XYZ_foo.
- Make mechanism_ppack_base header public.
- Tracing macros for modcc code gen.

* Add start time to `epoch` struct, rename time fields to avoid confusion with tfinal parameter to `simulation_state::run()`. Default constructed epoch now has id -1.
* Replace `t_`, `epoch_`, `min_delay_` timekeeping in simulation_state with `epoch_` and `t_interval_`: `epoch_` holds most recently simulated interval.
* Split event_lane population out of exchange task.
* Document in code comments the sub-task dependencies in `simulation_state::run()`.
* Rewrite main simulation loop to abide by pre- and post-conditions that ensure correct behaviour from successive calls to
`simulation_state::run()`.
* Replace double buffer for `simulation::local_spikes_` with array, analogous to `event_lanes_`.
* Update cell groups, unit tests for new epoch interfaces.
* Remove `util::double_buffer` and test.
* Remove unnecessary includes in `mc_cell_group.hpp`.
* Add unit tests for simulation global spike callback and restart operations (i.e. successive calls to `simulation::run()`).
* Fix for `lif_cell_group::reset()` error leading to out of bounds access.

Fixes #873 and #1426.

* Change implementation of current clamp from a builtin mechanism to
  functionality provided by the back-end shared state classes.
* Extend clamp functionality to cover more use cases: it now supports DC
  and AC stimuli governed by a piecewise-linear envelope.
* Update Python iclamp wrapper to suit; add unit test.
* Account for total applied stimulus currents separately, and remove
  their contribution to the reported transmembrane current and
  ionic (i.e. non-capacitive) transmembrane current.
* Add new probe specifically for tracking total current stimulus in a cell.
* Update docs, and add an entry for current clamps to the concepts doc.
* Remove entirely the builtin catalogue.

* Provide defines and variables for version string components: major
version; minor version; patch level version; development suffix.
* Change commit date format to strict ISO.
* Add define/variable for Arbor build configuration.
* Add usage/help info to `git-source-id` script.
* Add define/variable for a full build identification string,
incorporating source info, version, build configuration, feature
options.
* Add documentation for the version and build defines and variables.
* Extend version unit tests to suit.

* Remove parts of the C++ API which leaked into `doc/fileformat/neuroml.rst`
* Add links to the relevant C++ and Python API

* Correct some malformed YAML.

* Auto-generate binary Python wheels, triggered by git tags 'v*'.
  * Wheels are generated for Python versions 3.6-3.9, for Linux using the [PyPA manylinux-2014 image](https://github.com/pypa/manylinux) and MacOS using macos-latest (10.15 at time of writing).
  * The Action generates `sdist` for any other platforms.
  * The cibuildwheel action automatically prepares the wheels using [auditwheel](https://cibuildwheel.readthedocs.io/en/stable/options/#repair-wheel-command), which means bundling of external deps (e.g. libxml2 for NeuroML).
  * The output is all that's needed for publication of a new version to (Test)PyPI.
* `setup.py` defaults `neuroml` to on.
* A small change to `CMakeLists.txt` was required to build the wheels: based on this [Pybind hint](https://pybind11.readthedocs.io/en/stable/compiling.html#findpython-mode) `Development.Module`is now required, and not `Development` when CMake searches for Python. I confirm this change is needed to make Arbor build in the manylinux docker images, but I do not understand what the change implies, other than that we now don't link to `libpython`.

* Refactor Documentation layout. 
* Display Spack page.
* Display Extending catalogues page.

Fixes #1353.

Force nvcc to use the c++ compiler used to compile the c++ code.

* Fix a quite obvious ptr arithmetic bug.

* Clean-up to avoid similar issues.

* Const, my old friend.

* Test: refactor..

* Lambda capture.

* Clean up using append_chunk.

* Catch two escaped _.

Update `set_ion` method of `cable_global_properties` to allow the addition of new ions

* Remove redundant `initial_ion_data` specification.

* Use CMake check_cxx_source_compiles and check_cxx_source_runs to detect how std::filesystem
is enabled. The probes are executed first, if none succeeds, we use link-only to cover the possibility that we
are cross-compiling.

Preliminary work towards the mechanism ABI.

Unification of the `public` and `protected` interfaces of the mechanism classes.
These interfaces will be merged into the public API ('public' as in: exposed via `arbor/include/arbor`). 

Interface methods `nrn_` have been renamed
- `nrn_init`: `init` (would have preferred initialize, but that is taken)
- `deliver_events(event_stream)`: `apply_events` (to avoid a name clash)
- `nrn_state`: `advance_state`
- `nrn_current`: `compute_currents`

Additional changes:
  - Remove typedefs for index_type etc
  - Shuffle `ion_state_view` to public interface
  - Migrate `copy_extend` to a proper place out of `multicore/mechanism.hpp`. Maybe `memory` or `range_util`.
  - Remove `vec_t_to_`

* Update `README.md`: moved citation to `doc/index.rst`.
* Advertise Py3.9 compat PyPI in setup.py
* Fix typos.

* Tutorial ring network added. A few typos fixed, crosslinks fixed.
* MPI tutorial added.
* Improved SVG generation, added separate label scaling, increased default label size.
* Bugfix: didnt generate cable region in gen-labels.py used in `concepts/morphology.rst`.
* Termified Interconnectivity page.
* Termified probe nomenclature page. Will be moved in [this PR](https://github.com/arbor-sim/arbor/issues/1353) from python/simulation to concepts/probes_samplers.

`/doc/scripts/divio_docs_theme` has the contents of the tarball from PyPI, slightly modified (custom fonts removed, font-awesome reduced to woff2, logo enabled, divio banner/analytics removed.).

Remove transmogrify layer from s-expression parser.

The transmogrify layer isn't needed if asc files won't be parsed as s-expressions.
It complicates and slows down parsing of s-expressions.

- Allow SOLVE statement anywhere in the BREAKPOINT block.
- Add unit test.

Fixes #1384.

* add neuroML example
* no test, because GH actions arbor is built without neuroML support

* add support for windows new line in s-expression parser
* make small optimization that reduces time to read large strings by 20%
* fix another small bug with handling error states.

Fixes #1381

* Remove check for strict linearity in current contribution in modcc, as it is no longer necessary.

* Refactor the functions from `algorithms.hpp`:
  - `mean`: moved to `profile/meter_manager.cpp`
  - `make_index`: moved to `util/index.hpp`
  - `is_strictly_monotonic_increasing`: moved to `/backends/gpu/forest.hpp`
  - `has_contiguous_compartments`:  moved to `/backends/gpu/forest.hpp`.
  - `branches`: moved to `/backends/gpu/forest.hpp`.
  - `expand_branches`: moved to `/backends/gpu/forest.hpp`.
  - `tree_reduce`: moved to `/backends/gpu/forest.hpp`.
  - `is_minimal_degree`: moved to `tree.hpp`.
  - `child_count`: moved to `tree.hpp`.
  - `in_interval`: not used, deleted.
  - `is_strictly_monotonic_decreasing`: not used, deleted.
  - `all_positive`:  not used, deleted.
  - `all_negative`: not used, deleted.
  - `find_branch`: not used, deleted.
  - `is_unique`: not used, deleted.
  - `binary_find`: not used, deleted.
* Remove `make_parent_index` from `tree.hpp`: not used. 
* Replace `make_index` usages with `make_partition`.
* Add unit tests to `test_forest.cpp`, `test_tree.cpp` and `test_index.cpp`.

- Make reversal potential `e` a global parameter of `pas` to avoid inconsistency 
  (reversal potential is represented by a RANGE parameter, despite it not being a conductance multiplier see #1052)
- Adjust default value of `e` to -70mV (as in Neuron)
- Streamline docs,  which use `passive`, `el`, and `gl` instead of `passive`, `e`, and `g`.

Closes #1052

- remove repeated CMAKE_BUILD_TYPE
- bump lcov to 1.15
- use new --exclude flag to skip generated sources

With gcc 8 as a compiler + a new version of lcov (#1350) codecov started tracking generated source files and coverage dropped to < 60%. Not sure if this is a bug in lcov or not. I've added these new flags for lcov: `--exclude "*/ext/*" --exclude "*/test/*" --exclude "*/mechanisms/*"` to fix it.

Use TCP BTL for MPI in GitHub Actions to fix sporadic SEGFAULT on MacOS.

See open-mpi/ompi#6518

Adda Linux pipeline with SIMD and not MPI to GitHub Actions.

* Sync Cable Cell page names and ordering between sections (concept, python, cpp). Except cpp cable cell page. Is still one big page.
* Move CV definition across categories to consistent location.
* Move decor to separate page (cpp currently missing decor documentation) because it's a top-level concept that should be in the TOC like the other ingredients of the cable cell.
* Slight rework of Concept intro and Cable Cell pages.
* The Cable Cell page is now basically an intro, and lay out how the other pages/concepts relate to each other.

Online STDP is implemented in the synapse and
and demonstrated by recording an STDP curve in
an example.

* fix nmodl comment parsing
* proper handling of \r
* fix segmentation fault when api_state is null
* add unit tests