[ENH] common mixin for `BaseDeepClassifierPytorch` and `BaseDeepRegressorTorch`

Is your feature request related to a problem? Please describe. There are many common functionalities between the classifier and regressor base classes, and this refactor would enable to move all the common methods to a common mixin. This will help keep only the necessary logic in the base classes. This will also make it easier for new contributors to build new estimators on top of the classes. This will also help to resolve bugs in a single place rather than having to resolve it at multiple places each time, which is redundant.

Describe the solution you'd like A common mixin that is used by both the classifier and regressor base classes. The main aim is reusing common functionality.

Reference Issues and PRs

• https://github.com/sktime/sktime/issues/10306
• #11151
• https://github.com/sktime/sktime/pull/9390#issuecomment-4583625335