Store analysis results / debug info for benchmarks

[adonath]

To be able to compare benchmarks we should make sure the analysis results stay the same. E.g. to understand whether a change in performance comes from improved code or just a different number of iterations for a fit. String the results in YAML somewhere and just compare visually with git diff should be fine for now.

[adonath]

Also we should store the datasets after reduction and fit traces...