estimate_ate in Xlearner triggers a retrain of the model when pretrain=True and external propensity scores are used

[AlxClt]

Describe the bug
When calling BaseXlearner.estimate_ate() with pretrain=True and external propensity scores (i.e. p is not None), a retrain of the model is triggered because the case where pretrain=True and p is not None is not handled

To Reproduce
Train the Xlearner with externally computed propensity scores and call estimate_ate using pretrain=True and external propensity scores. The model will be retrained.

Expected behavior
The model should not be retrained, instead the ate should be computed using the passed propensity scores

Screenshots
If applicable, add screenshots to help explain your problem.

Environment (please complete the following information):

• Windows
• Python Version: 3.8.10
• pandas==1.2.4, scikit-learn==1.2.0, cython==0.29.23

Additional context
Add any other context about the problem here.

[AlxClt]

Fix should be very simple, working on it

[AlxClt]

Fix in PR #597