Normalise runs and sessions names

[PierreGtch]

• Closes Normalise runs and sessions names #451
• new naming convention: str of an integer corresponding to the run/session recording order (starting at 0), eventually followed by a description string (only containing letters and integers but not starting by an integer)
• Adds a test in dataset.get_data that verifies if the runs and sessions and subjects are correctly named

[PierreGtch]

This comment will be removed from the code but is relatively important:

In the BIDS standard, the runs are expected to be indexes. This is not always the case in MOABB because we allow to append a description to the index. See: https://bids-specification.readthedocs.io/en/stable/glossary.html#run-entities

[bruAristimunha]

I liked it, but I'm unsure if this won't break something in the braindecode. Maybe we need to do something more activate. Add a little sub test routine or something else.

[bruAristimunha]

I will review the PR more carefully

[PierreGtch]

Let’s normalise session and run suffixes to “test” and “train”.
For the special cases (soa, Cattan2019_VR), we keep them like that for now.

[sylvchev]

Thanks @PierreGtch this is really a nice API breaking change ^^

[sylvchev]

closes #472

[PierreGtch]

This is ready to merge @bruAristimunha @sylvchev

[bruAristimunha]

Amazing work @PierreGtch. I need to pay you a beer =)

[gcattan]

Missing update in bci_illiteracy:

[PierreGtch]

Indeed, thanks @gcattan. I made a PR: #511