Théo Gnassounou

This PR allows the user to put several layer names to adapt to different network levels, as done in several papers. This PR also forces the features of each layer...

I like the one of braindecode (https://github.com/braindecode/braindecode/blob/master/CONTRIBUTING.md) Having one can help people contribute. I see a lot of people struggling with pre-commit, for example.

[TO_REVIEW] fix `predict_proba` for deep method

2

The `predict_proba` of deep method was choosing the nonlin attributes by looking at `criterion` attribute of the solvers. The `DomainAwareNet` has a `DomainAwareCriterion` as criterion. We should choose the nonlin...