Add Green model to benchmark - follow up

[jpaillard]

Following up on #39 this integrates the review from @tomMoral

[tomMoral]

LGTM! thanks for the follow up.

Before merging, can you share a screen shot of the result on BNCI competition with one competitor like CSP-LDA?

[jpaillard]

I ran this yesterday against the "Aug-Cov_reg-Tang-SVM". The time may not be so informative as it ran on CPU.

[tomMoral]

The neuro-green package seems to cause some issues due to pinned dependencies.
I had to skip its install in #13 so it does not break the tests for others

[jpaillard]

Could you indicate the problematic dependencies?

[tomMoral]

It is not completely clear as I could not pinpoint exactly which dependency caused the issue but in #13, skipping green install solved my issues.
I will try to see if it is something else (I just added back the install) but if not, we need more investigation.
If possible to avoid having fixed dependencies (with ==), that would help a lot.

[tomMoral]

Ok it seems that the error was mostly bad interactions with another error with the latest version of scikit-learn.
So it is now working with both #13 and green.
That being said, it would be great to unpin the dependencies as this can break environments easily (typically this downgrade numpy<2 while some packages might require it to be above).

[jpaillard]

I tried to solve this in Roche/neuro-green#6, hope it's better now.

[tomMoral]

Seems much better! thanks @jpaillard :)