Splitter

[thibaultdvx]

The main goal of this PR is to make the splitter module ready to use. To do that, I did:

• adapt the splitters to the new ConcatDataset, PairedDataset, and UnpairedDataset;
• enhance docstrings;
• create pages on the splitters in the documentation;
• update test data, to be consistent with the other tests;
• complete unittests.

I also did:

• change a bit the subset function of CapsDataset (and its derivatives), so that it doesn't raise an error if a (participant, session) couple in the input DataFrame is not in the datasets;
• remove unused functions from split_utils.py, and put this file in make_splits because it only concerns make_split and make_kfold;
• rewrite some tsv tools like read_data (function to read a tsv) in tsvtools/utils.py, and remove unused functions in this python file;
• remove default values from SingleSplitConfig and KFoldConfig to only keep them in make_split and make_kfold;
• put the checks made on the split directories directly in SingleSplitConfig and KFoldConfig;
• reduce docstrings of private functions in order to have smaller files;
• reorganize files to have always the main function at the top.

Besides, I would particularly want to have your opinion on two changes I made:

• for SingleSplit, I suggested to change get_splits to get_split, because I found it confusing to have plural here;
• I also changed validate_longitudinal to longitudinal.

[thibaultdvx]

To fix multiprocessing issues with num_workers for Python 3.12 on MacOS, I made pytest skip the multiprocessing tests on MacOS machines.

[camillebrianceau]

LGTM! Let's merge it so I can rebase before Monday!

And you can change get_splits to get_split if you prefer.