- Add of benchmark scripts that compares perf wrt pygmalion.
=> As expected it is bad in terms of reading cost (because of row by row
matrix read) but it is also not good in terms of simulation cost.
hcat seems to perform badly, should investigate.