Decoy Implementation and Additional Selection Methods in Snekmer Learn/Apply

[jjacobson95]

This PR brings Snekmer Learn/Apply up to date with numerous developments that have occurred since the last update.

These updates include:

• Reverse Sequence Decoys
• Family Specific Thresholds
• Additional Selection methods (such as 'greatest_distance')
• Added an output directory from Learn called "apply_inputs" that consolidates the 3 key files needed for Apply.
• Additional misc bug fixes.

To do: use snakefmt for linting - I did a quick test and it broke formatting, so I'll come back to this when we want to merge the PR.

---

Additional possible updates in this or a future PR include:

• Many to Many 'Apply': allow users to use sequences for hits instead of families.
• Inclusion of values below the decoy threshold. Currently these are excluded, we should instead add a flag and let the user decide if they want to use them.
• Multi-hit function: Allow there to be multiple hits for a query sequence. This may be done by utilizing the family specific cutoffs and allowing all scores above a threshold to pass.