Annarchy

* Added more unittests, e. g. spike-based plasticity rules, IndividualNeuron accessors
* Added sanity check that no duplicate entries occur in list provided to visible_cores
* Bugfixes:
* Fixed code generation errors (eb6d3db, eb6d3db, 8eb80ee, cf7e9e6)
* Fixed a bug in connector methods

* Added new sparse matrix format: sliced ELLPACK (sell)
* 4th-order Runge-Kutta is now available as numerical method (`rk4`).
* Improved existing sparse matrix formats:
* blocked sparse matrix format (BSR) now implements the complete interface (60d15cc)
* added a modified DenseMatrix specialized for PopulationViews (fcba8bd)
* for matrices using dense formats (ELL, ELLR, Dense, BSR) fixed a bug when the sanity check was called too frequently (7f75c8b)
* Added isfinite() and isnan() to usable keywords in equations/spike conditions
* Added *sparse_matrix_format* to setup() arguments to globally set the used sparse matrix representation for Projections
* Added *annarchy_json* argument to parallel_run()
* Improved profile:
* adding --profile will provide a performance overview (on console) when the script ends.
* extended the number of functions measured for ANNarchy models run on GPUs.
* Improved code generation for single precision use-case: constants and math functions (sin, cos ...) are extended by a f-suffix
* Improved performance for the spike_gather operation on GPUs.
* Improved sanity checks for not implemented features.
* Changed the Python type of Population.ranks from list to numpy.array to reduce memory footprint (5f9d58b , ff9c934, fe5b937)
* Extended the applicability of thread_placement (see also visible_cores argument to setup): useable now in context of parallel_run()
* Allow the usage of a psp-term in PoolingProjections.
* ANNarchy executed with -v/--verbose notifies now also about changed files (helpful to track down why ANNarchy recompiles)
* Unittests (Thanks to Alex for his contributions on that!)
* revised the structure of the unittests
* added possibility to cover sparse matrix formats aside of LIL / CSR
* added unittests for functions like save/load of parameters
* added new unittest for PopulationViews (2893199)
* Bugfixes
* fixed a replacement error made by the parser when targets had overlapping names (8df6eb7)
* fixed code generation bugs related to non-uniform delays (064f68d)
* fixed access methods for individual synapse entries for (ELLR, CSR)
* spiking models: CSR, CUDA (c5c20b3, cd574fa); Dense, CPU (8ef4d0b); CSRC, CPU (4dd0aaa); CSR, CPU - in context of PopulationViews (6485149, 4f9c7db)
* updating the fiters for convolutions after compile() was not possible (a8f78d1)

* Bugfix: indices used by IndividualSynapses.
* Bugfix: calling simulate() after a Monitor.stop() call crashed.
* Bugfix: period_offset was ignored when creating copies of Monitor objects.
* Bugfix: wrong indices used in CSRC_T in case of PopulationViews.

* Bugfixes:
* In rare cases the 'w' variable could not be recorded.
* Bugfix in early stopping of simulations (simulate_until).
* Bugfix in CUDA code generation using dense matrix format.
* Bugfix in code generation related to Runge-Kutta 4th order method and refractoriness.
* Bugfix in cython wrapper for Monitor (github issue 3)

* `reset()` has a new argument *monitors* to control the clearing of monitored results.
* Improved error messages for non-implemented features on CUDA devices.
* Bugfixes:
* fixed bugs related to compuation of mean-firing rate for spiking models.
* fixed a bug related to IndividualNeuron accessor.

* Extended the available set of hand-written SpMV codes (in particular SSE and AVX-512)
* Bugfixes:
* the usage of storage_format="dense" and a constant weight (e.g. in the COBA example) leads to an error message.
* Network.enable_learning() had no period/offset arguments.
* TypeError occured in connect_fixed_number_post() cause by a type mismatch on a temporary variable.