Feature/bidirectional lstm decoder

[lard2forecast]

Description

This feature implements the possibility of using bidirectional lstm decoder layers in the Temporal Fusion Transformer model. There are some additional notes in the code, which should be considered during code review.

Checklist

• Linked issues (if existing)
  No existing issues, but possible caveats:
  https://discuss.pytorch.org/t/proper-way-of-setting-h-c-in-bidirectional-rnn-layers/158842
• Amended changelog for large changes (and added myself there as contributor)
  Changes too small to be of significance in a changelog (i.m.o.)
• Added/modified tests
• Used pre-commit hooks when committing to ensure that code is compliant with hooks. Install hooks with pre-commit install.
  To run hooks independent of commit, execute pre-commit run --all-files

Make sure to have fun coding!