Is your feature request related to a problem? Please describe.
The EncoderDecoderTimeSeriesDataModule (D2 layer) currently does not apply feature scaling (for continuous variables) as part of its preprocessing pipeline.

Describe the solution you'd like
The scalers can be implemented within the _preprocess_data method and called during setup. We fit the scalers on the training data and then transform each individual batch inside the _ProcessedEncoderDecoderDataset.

Additional context
#1974 - Work items for PTFv2.