Add profiling to tsfresh

[Scott-Simmons]

We could add profiling as a make target and/or script. We could take the reprexes from open issues and use their data to profile.

It could be a good way to branchmark subsequent features/improvements to tsfresh. Then can start tackling some of the issues below with some more rigour

• speed up calculation by numba #205
• Feature extraction progress bar fills fast, but some part in the code takes a long time #703
• Parallel computing not only for Distributor, but also for DataFrame-related preprocessing #728
• Even 512 GiB Memory not enough for extract_features on 7895800 rows × 28 columns ?  #947

(jumping the gun a bit, but I imagine that we can find some low hanging fruit somewhere to improve performance, but first having a measuring framework could be a good start)