refactor timestamps managing

[Gautzilla]

This PR follows the changes made regarding issue #209

[Gautzilla]

@mathieudpnt @cazaudo, basically this PR does:

A rework of how the timestamps are parsed from the audio file names. I added some basic utils which might simplify the old write_timestamp function:

• Dataset now uses the _write_timestamp_csv_from_audio_files function to write the timestamp csv from the audio files present in the audio folder. this function:
  • get the supported audio files from the audio files folder
  • parse the Timestamps from the audio file names thanks to the provided format pattern string (this is the step that uses all the added utils, see associate_timestamps in the utils\timestamp_utils module)
  • write the timestamps.csv with the uniformed OSmOSE datetime string format (defined in the config module)

I added a bunch of unit and integration tests to make the thing robust, and it should not break the global (untested) OSEkit workflow since the timestamp.csv is written in the same fashion as before (except for the timestamp precision, which has been reduced to the millisecond).

I intentionally kept the build() method as it is for now (and integrated the latter changes to it), but these changes might help refactoring the build() method (I think this is the next step!)

The utils added to timestamp_utils should also be used elsewhere in the codebase (e.g. by the reshaper), I'll do that in the near future too.

PS: pytest throws some warnings on the Python 3.10 used in the github actions here. I managed to solve them locally in an environment that uses Python 3.12.7: is there a reason for sticking with the 3.10 version?

[mathieudpnt]

a few minor changes, overall looks good to me!