Adding SPARCNet and ContraWR

[bruAristimunha]

In Biot's article, in addition to the BIOT model, the author centralizes codes from other models.

The list is:

1. SPaRCNet,
2. ContraWR,
3. CNN-Transformer,
4. FFCL and
5. ST-Transformer (predecessor of EEG Conformer, which we already have implemented here).

In this PR, I am proposing to implement 1) SPaRCNet and 2) ContraWR.

3. CNN-Transformer seemed very similar to ContraWR, I don't think it's worth implementing.
4. The FFCL does not appear to be correctly implemented.

[bruAristimunha]

fyi @mccorsi

[codecov]

Codecov Report

Attention: Patch coverage is 99.22481% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 87.20%. Comparing base (77b4f8b) to head (b1caba7).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #611      +/-   ##
==========================================
+ Coverage   86.93%   87.20%   +0.26%     
==========================================
  Files          65       67       +2     
  Lines        5796     5925     +129     
==========================================
+ Hits         5039     5167     +128     
- Misses        757      758       +1     

[agramfort]

at some point it would be good to do a full evaluation of all these architectures on a some standard datasets

[bruAristimunha]

The green button is yours @agramfort