TargetEncoder should respect sample_weights

[mayer79]

Describe the workflow you want to enable

The current implementation of TargetEncoder seems to calculate (shrinked) averages of y. In cases with sample_weights, it would be more natural to work with (shrinked) weighted averages.

Describe your proposed solution

In case of sample_weights, shrinked averages should be replaced by corresponding shrinked weighted averages.

However, I am not 100% sure if sample_weights are accessable by a transformer.

Describe alternatives you've considered, if relevant

The alternative is to continue ignoring sample weights.

Additional context

No response

[betatim]

In principle sample_weight can be accessed when fitting TargetEncoder. So at first sight this is something that can be considered as a new feature. There seems to be some support for it as well (👍 on the issue).

[ogrisel]

Related to #11316 for testing.