Add mapping-tester repetitions as separate script

[fsimonis]

Main changes of this PR

Based on #225
Replaces #195

This PR adds a separate python script repeat.py to the mapping tester to handle repetitions, which are useful for runtime measurements.

It repeats for each repetition:

1. runall
2. postprocessall
3. gatherstat into file-n.csv

Then it merges all N file-n.csv into a single file.csv.
The column run indicates which repetition it was.

Author's checklist

• I used the pre-commit hook and used pre-commit run --all to apply all available hooks.
• I added a test to cover the proposed changes in our test suite.
• I updated the documentation in docs/README.md.
• I added a changelog entry in ./changelog-entries/ (create if necessary).
• I updated potential breaking changes in the tutorial precice/tutorials/aste-turbine.

[davidscn]

There is currently no option to recover details about individual runs, once all repetitions are completed, right? Or do we store the json file?

[davidscn]

I think with this we should also provide a script to process the data statistically across various runs, no?

[fsimonis]

The script currently doesn't clean up intermediates. Running it for 5 repetitions results in:

test-statistics-0.csv
test-statistics-1.csv
test-statistics-2.csv
test-statistics-3.csv
test-statistics-4.csv
test-statistics.csv

The various profiling.json files are not stored.
Isn't it easier to integrate the interesting events directly into gatherstats?

[davidscn]

Ah ok. I think then this essentially a python version of what I once implemented in bash. The nice thing about the bash script was actually that it served a central steering component around all the scripts in the mapping tester. However, you can still add the repetition script provided here, as one can still use the bash version as a central instance if deemed necessary

[fsimonis]

Added a simple aggregation script for mean median and variance with support for dropping min and max.
I integrated it into the runtime tests, even though the column values will not be checked.

Will add this as a separate PR. Adding backends for both polars and pandas makes this a bit painful.