Refactor benchmarking scripts and update generated plots

[denisalevi]

Make benchmark scripts easier to use and update the plots that are automatically generated, e.g. add plots form paper.

• Get rid of benchmark related things in the Brian2 patch. Either add to Brian2 or move it all into Brian2CUDA repository.
• Refactor the run_benchmarks.py, such that config / benchmarks / sizes configuration is separate (and clean and tidy) from the actual run script. Maybe just a config file for that? And a parameter to the bash script that chooses the config file?
• Add the plotting code from the paper to the brian2CUDA repo and adapt the automatically generated plots. The plots should also be chosen in the config file? Or everything is always plotted?
• Add an option to e.g. only compare stateupdaters across configurations with a simple bar plot? Or maybe the profiled bar plots from the paper are already good enough?
• Change DynamicConfigCreator to automatically generate the config name to avoid different names for same configs in the future...