Make `_drop_time_index()` public and call when using `transform_all_but_final()`

[christopherbunn]

• As a user of EvalML, I expect an estimator-ready dataframe when I call transform_all_but_final(). However, for time series problems, this dataframe includes the datetime column even if it is dropped before calling the estimator. Thus, the time index column should be dropped when transform_all_but_final() is called.
• In addition, as a user of EvalML, I would like to have a built-in pipeline function that I can call that drops the time index column. This exists as _drop_time_index(), but is not meant to be called publicly.