Describe the workflow you want to enable

I would like to propose an enhancement to the check_estimator function, particularly in light of the implementation of SLEP007. As per SLEP007, which was integrated into scikit-learn starting from version 1.1.0, estimators are expected to support feature_names_in_ and n_features_in_ attributes. However, it appears that the check_estimator utility does not currently enforce the presence of these attributes.

Describe your proposed solution

I propose that the check_estimator function be updated to include checks for the existence of the feature_names_in_ and n_features_in_ attributes.

Additional context

• SLEP007: https://scikit-learn-enhancement-proposals.readthedocs.io/en/latest/slep007/proposal.html
• Version of scikit-learn where SLEP007 was implemented: 1.1.0
• Example of check_estimator not enforcing these attributes: [python-package] Support feature_names_in_ attribute via sklearn API microsoft/LightGBM#6279