Normalise runs and sessions names

[PierreGtch]

The naming scheme of the sessions and runs is not always consistent between datasets. They are usually prefixed by "session_" and "run_" but not always. They are usually followed by a number but not always (ex. Schirrmeister2017 has runs names "train" and "test"). If it's a number, it sometimes starts at 0 and sometimes at 1.
(The subjects names are always integers and, I think, always starting at 1)

The goal of this issue is to normalise all this to have run and session names that always reflect their chronology.
And while we are at it, we could directly make these names compatible with the BIDS specifications (i.e. only with letters and numbers. No spaces, dashes or underscores).

So we could use the following guidelines for (re)naming:

• remove the prefixes session_ and run_ for compatibility with BIDS,
• always begin the name with an integer that tells in which order the session/run was recorded (starting at 0 or 1?),
• eventually add a string after the integer to provide additional information (ex: train/test) without separation in-between.

Also, we should add tests when loading the data and at least raise a warning if the names are incorrect.