Matrix plot and Track Origin confusion matrix

[LucioDerin]

Summary

This pull request introduces the following changes

• Added the class MatshowPlot dedicated to matrix plots, based on matplotlib's matshow (can be found in puma.matshow).
• Added the function confusion_matrix to compute a confusion matrix from target and predicted labels of a classifier. Can be found in puma.utils.confusion_matrix.
• Added the confusion matrix plot for the Track Origin Auxiliary Task, as a method in the AuxResults class. The result on the dummy_GN2 tagger is attached at the end.

Relates to the following issues

• Mildly related to issue #107, as the MatshowPlot class can be used to plot the saliency map once it's computed (a similar plot, obtained with the MatshowPlot class and implemented in the examples, is attached at the end).

Conformity

• Changelog entry
• Documentation

Examples

Track Origin Confusion Matrix

A general matrix plot (inspired to the saliency map)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.89%. Comparing base (830a27f) to head (2caf36b).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #258      +/-   ##
==========================================
+ Coverage   97.81%   97.89%   +0.08%     
==========================================
  Files          43       47       +4     
  Lines        4661     4852     +191     
==========================================
+ Hits         4559     4750     +191     
  Misses        102      102              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dkobylianskii]

Thanks a lot for this MR, it looks really good.
I've added some minor comments

[dkobylianskii]

Thanks again for implementing this, @LucioDerin