Add Python profiling tools

[dtch1997]

Comments:

• As well as pyinstrument (which does sample based profiling so doesn't record all stack frames only a sample) there is the built in cProfile module (https://docs.python.org/3/library/profile.html) which deterministically profiles all function calls - this gives more complete coverage but at a higher overhead (and the overhead in some cases may make the profiling timings non-representative of the non-profiled code). If you go with cProfile then snakeviz (https://jiffyclub.github.io/snakeviz/) is often a nicer way to visualize the profiling output than the default tabular output

Resources:

• https://github-pages.arc.ucl.ac.uk/python-tooling/pages/benchmarking-profiling.html
• https://docs.nersc.gov/development/languages/python/profiling-debugging-python/
• https://github.com/tkoskela/hpc_lecture_notes/blob/master/cluster_challenge/06_profiling.md