Trained random forest models should have a .plot_feature_importances() method

[Aylr]

While this would benefit from #446 it could be done with a bit of hacky logic as an MVP

GWT

• Given I am a user that has a trained random forest model

• When I call the method trained_rf.plot_feature_importances()

• Then I should see the feature importance plot shown during training.

• Given I am a user that has a trained model of any other type

• When I call the method trained_foo.plot_feature_importances()

• Then I should see a warning or error message saying that this is only for random forest models

Notes

• The current implentation of plot_rf_features_from_tsm() requires the x_train data. This isn't something we want to store in a serialized model for PHI, data security and performance reasons. Therefore, this will require some thought and investigation.
• Some possible solutions:
  • Generate the plot and serialize it somehow during training as part of the TrainedSupervisedModel
  • ???