cache_predictions does not compute all possibilities in EstimatorReport

[glemaitre]

cache_predictions does not compute all the predictions that should be cached in EstimatorReport. It should be the cartesian product of the parameters

The underlying bug is in:

skore/skore/src/skore/sklearn/_estimator/report.py

Lines 162 to 179 in 5314915

	if self._ml_task in ("binary-classification", "multiclass-classification"):
	if response_methods == "auto":
	response_methods = ("predict",)
	if hasattr(self._estimator, "predict_proba"):
	response_methods = ("predict_proba",)
	if hasattr(self._estimator, "decision_function"):
	response_methods = ("decision_function",)
	pos_labels = self._estimator.classes_
	else:
	if response_methods == "auto":
	response_methods = ("predict",)
	pos_labels = [None]
	data_sources = ("test",)
	Xs = (self._X_test,)
	if self._X_train is not None:
	data_sources = ("train",)
	Xs = (self._X_train,)

We redefine some tuples using = instead of concatenating using +=.

We should add a non-regression tests to check the what is getting cached once we call the function.

[augustebaum]

Can you be more precise? This way someone else might be able to work on this.

[glemaitre]

Sorry about that. It was a reminder to myself and thus I self assigned as well. The bug fix is in the cross validation PR. I might extract it out in another PR.