* SimulationConfig: Adds "represents_physical_electrode" field in "inputs" section

By using an `std::array` we can remove the differences between
`RawIndex` and `Selection::Ranges`. It also allows us to read
`Selection::Ranges` from disk directly via the HighFive API.

This change breaks API. One change needed is:
```
- ranges.emplace_back(i_begin, i_end);
+ ranges.push_back({i_begin, i_end});
```
another is:
```
- range.first < range.second;
+ std::get<0>(range) < std::get<1>(range);
```
both are valid for versions before and after this commit. Python code is unaffected.

* test for completely empty datasets

* prevents failures with cryptic error message:
    HDF5-DIAG: Error detected in HDF5 (1.10.4) thread 140737340061504:
    #000: ../../../src/H5Shyper.c line 7057 in H5Sselect_hyperslab(): hyperslab doesn't support H5S_NULL space

* Implement merge-read-extract.

Reading from parallel filesystems, e.g. GPFS, requires reading few but
large chunks. Reading multiple times from the same block/page, come with
a hefty performance penalty.

The commit implements the functionality for merging nearby reads by
adding or modifying:

  * `sortAndMerge` to allow merging ranges across gaps of a certain
    size.

  * `bulkRead` to read block-by-block and extract the requested slices
    in memory.

  * `_readSelection` to always combine reads.

  * `?fferent_edges` to optimize reading of edge IDs.

It requires a compile-time constant `SONATA_PAGESIZE` to specify the
block/pagesize to be targeted.

* Move include.

* Add missing reversal property of shotnoise family
* Explicitly set default value for 'reversal' field

* Extend metadata type to be int, float, bool and string

ie:
    vasculatureFile
    vasculatureMesh
    endfeetMeshesFile
    microdomainsFile
    spineMorphologiesDir

* SimulationConfig: random_seed in inputs default to None

* switch to having `connection_overrides` being a list
* required so that order is guaranteed: see https://github.com/BlueBrain/sonata-extension/pull/24

* it's common to have a deep structure of compound statements
   (ie: a whole hierarchy of regions), all checking the same attribute
   rather than `materializing` them separately, we group them, and materialize
   them all at once
* Reduces a full brain hierarchy from 40 minutes to 6 seconds.

* instead of Multi Clauses like being treated at the top level
    {
        "population": "biophysical",
        "node_id": [1, 2, 3, 5, 7, 9, ...]
    }
    They are converted the NodeSetBasicMultiClause
* this is in prep for faster `materialize`

* SONATA sim config: change the default filename for report

see https://github.com/BlueBrain/sonata-extension/pull/17 for semantics.

* SONATA sim config updates

See:
https://github.com/BlueBrain/sonata-extension/pull/19

Remove key "synapses_init_depleted" in "conditions" section (BBPBGLIB-968)
Move key "spike_location" from "run" to "conditions" (BBPBGLIB-914)

Without command `catch_discover_tests` is not defined when not using
vendored Catch2.

Documentation suggesting we should have that line:
  * (devel) https://github.com/catchorg/Catch2/blob/devel/docs/cmake-integration.md#automatic-test-registration
  * (lastest) https://github.com/catchorg/Catch2/blob/v2.13.10/docs/cmake-integration.md#automatic-test-registration
  * (vendored) https://github.com/catchorg/Catch2/blob/v2.13.6/docs/cmake-integration.md#automatic-test-registration

It also suggest that if Catch isn't located in a standard location one should
edit `CMAKE_MODULE_PATH`. However, doing so in `extlib/CMakeLists.txt` wont
propagate the change to `tests/CMakeLists.txt`. Therefore, we conditionally
include Catch instead.

* fixes #237, found by @1uc 

* circuit_config.json must now conform to the BBP standard for
  populations to be found
* the circuit_config.json now is the ground truth for available
  properties: if multiple nodes files contain the same population name,
  only the one referenced in the circuit_config.json is considered

* SimulationConfig parser: add modifications properties in conditions section

* remove forward_skip key from run (BBPBGLIB-913)
* remove minis_single_vesicle key from conditions (BBPBGLIB-915)
* add keys for additional seedings (BBPBGLIB-920)
      stimulus_seed, ionchannel_seed, minis_seed, synapse_seed
* add series_resistance key in seclamp (BBPBGLIB-921)
* add keys neuromodulation_dtc and neuromodulation_strength in connection_overrides
(BBPBGLIB-885)

Co-authored-by: Mike Gevaert <michael.gevaert@epfl.ch>

* remove the .getJSON method; one can just load the open themselves,
  MANIFEST resolution is the hard part

* adds relative_shot_noise, absolute_shot_noise, ornstein_uhlenbeck, relative_ornstein_uhlenbeck

* Parse connection_overrides section

* This way the chance of using the incorrect parameter is reduced, as errors will be created on access

* makes getting names to use for getReport/getInput simpler
* fixes #198

* Parse 'output' and 'conditions' sections in SimulationConfig

Co-authored-by: Weina Ji <weina.ji@epfl.ch>

* Add parsing for the inputs section
* Parse mandatory "random_seed" in the run section