Gabriel Daiha

Avoid double impurity calculation inside criterion.pyx in tree module

1

Inspired by https://github.com/scikit-learn/scikit-learn/pull/18203/files# in Sklearn`s RandomForest, this PR aims to avoid double calculation of impurity inside criterion.pyx of tree module

Add support for customize scoring methods in DRLearner, similar done to DML. This is a remake of #1000 to avoid problems with not signed-off commits.

Add support for sample_weights in DRTester

2

To solve issue #1004 , this PR adds support for sample_weights in DRTester class. It implements it in fit_nuisances and in each evaluation (calibration, uplift and blp).

I work with summarized data to make causal inference, where each row corresponds to a group of units. When I try to use DRTester to validate results, the methods inside...