[ENH] Probabilities calibration

[annapintara]

Scores returned by predict_proba method after applying undersampling or cost-sensitive learning are biased - they don't equal probabilities and need calibration.

However, there is a simple transformation to calibrate such scores, obtained from Bayes' rule, as described in the following article: https://ieeexplore.ieee.org/abstract/document/7376606 (over 700 citations). It would be great to implement this method.

[glemaitre]

I assume that it is the transform that is applied in this PR: #1077

An alternative is to apply one of the calibration method: https://scikit-learn.org/stable/modules/calibration.html