Add, standardize and document DART outputs

[oliche]

Spike sorter have sloppy, non-complete and non-documented output formats.
Transformation to SI units is lost, some useful datasets are omitted and quality control non-existent.
Here it is the opportunity to write down the outputs of this spike sorter and specify some qc metrics along the way.

• preprocessing: ap rms, spectrogram, bad channels, raw data snapshots (raw / destripe / residuals)
• localisation hdf5 or parquet dataframe
• drift: vertical shifts array, snapshots
• spike sorting: spike trains as usual, sample waveforms, templates

Note 6/29: HDF5 files in particular can have some issues with parallel processing e.g. https://stackoverflow.com/questions/68457924/error-while-accessing-hdf5-file-shows-error-oserror-unable-to-open-file

[oliche]

For now this is applicable only to drift and pre-processing