Note: failed assertion at run time with both fvm_cell
and fvm_multicell.

Latest Multicore Optimizations

Minor fixes for PR#86 branch.

* Count tests with int not bool in `mpi_listener.hpp`
* Ytpo fix in `gathered_vector.hpp`
* Add missing include in `test_algorithms.cpp`

* fix compile time error in src/model.hpp
* add mpi_listener type for sensible output and error reporting
  when using MPI with Google Test.
* add stub for communicator unit testing

Migrate cell_group and fvm_cell to internal API.

* Initialize voltage in `fvm_cell::initialize` to `resting_potential_`.
* Remove errant end-of-line whitespace.
* Fix merge conflict (merged master).

* fix build errors in tests with TBB
* set default probe output to none in miniapp

* the user can specify one of {none,AVX,AVX2,KNL} as
  targets for vectorization
* if modcc is available in PATH it will be used, which
  makes cross compilation easier
* a new flag USE_OPTIMIZED_KERNELS can be set to
  generate optimized kernels from the modcc compiler
  (equivalent to adding -O flag to modcc)

* `nest::mc::threading::sort` is analogous to `std::sort`
    * when using TBB the `tbb::parallel_sort` is used
    * otherwise `std::sort` is used
* used to parallelise connection construction during setup
    * now all initialization and simulation is multi-threaded
* added a unit test

* replace a call to `std::iota` with a for loop because the Intel
compiler was optimizing out the `std::iota` call in a unit test.
* add uniform syntax for converting pointer+size and ArrayView+size
into an ArrayView to provide uniform method for accessing ranges
inside mechanisms with/without optimizations.

Ranges and views implementations.

* Makes EXPECTS() fail handler a global, settable function pointer.
* Fix `util::upto` for bidirectional iterators with begin==end.
* Use `std::string` in error reporting with `either` in `partition_view`
  to correct incorrect default conversion in return value.
* Correct return types for `util::range::operator[]` and `util::range::at()`
* `transform_iterator` requires only iterator==sentinel test for correct sentinel
  behaviour.
* Extend partition tests to cover:
  * partition views over short (length) sequences,
  * non-numeric partitions,
  * throw on validation check of non-monotonic sequence,
  * make_partition with short and long size sequences, short division containers.
* Extend range tests to cover:
  * compatibility with `std::accumulate`,
  * correct constness of dereferences with respect to wrapped iterators,
  * proper type deduction for `make_range` over pointers.
* Extend span tests to cover:
  * span creation from `std::pair`,
  * proper type promotion with heterogeneous bounds.

* make a new gathered_vector type for storing a "distributed" vector
  that has been gathered onto a single node
    * stores the vector
    * stores the partition information
* split communicator::exchange() into two phases
    * exchange() which simply exchanges spikes
    * make_event_queues() which generates local queues from spikes
* now callbacks for both local and global spike store can be
  performed in the model
* and extra meta data about the partition is available for more
  advanced parallel global storage schemes.

* Adds non-owning range class describing a half-open interval
  by a pair of iterators, or by a start iterator and sentinel.
  (Assumption is that these are bounded ranges; unbounded
  ranges are on the todo list.)
* Adds spans and counting iterators, for iterating across
  intervals of integral types with a range-like interface.
* Partition in-place creation functions and partition views
  over monotonically increasing sequences.
* Two-field variant type implementation `util::either<A, B>`.
  (Used to provide an STL- and range for-loop accessible
  façade to sentinel-terminated ranges.)
* Some simplifications of utitliy classes and tidying up
  of formatting.

* Define an internal API for lowered cells.
* Devolve all responsibility for index
  translations to `cell_group`.
* Reduce state/info required in lowered cells,
  for example probe information and cell segment
  indices.
* Fix (to confirm) issue where a cell could have
  synapses of only the one type.
* Spike detectors now keep a lowered cell-specific
  handle instead of a segment location.

Global and Local spike exchange streamline

Exporter fixes

* single file per rank output creates one file per rank
* global file output creates one file only from rank 0

* MPI ranks now test the file that they generate, not that rank 0
generates
* code formatting updates
* added interface for getting file file name used by exporter

* single flag -f that turns output on (off by default)
* other parameters have to be provided by an input file
    * if -f is set the defaults will be used