Casanovo-DB potential performance improvements

[bittremieux]

• Should give a warning when non-supported PTMs are used (e.g. C carbamidomethylation is not specified as a fixed modification) [easy]
• Multithreading during FASTA digestion and m/z calculation [easy]
• Refactor candidate selection to not search from scratch for every candidate but use some sort of looping index [moderate (within batches) — hard (across batches, would require modifications to the data loader)]
• mzTab export should report protein database information [easy]
• Spectrum progress bar should show more granular updates [moderate]
• Superfluous predicted peptide m/z calculation (can be derived from database) [moderate]

Open-ended evaluations:

• Profiling to understand where the runtime and memory consumption bottlenecks are—likely PSM batch creation contributes. Candidate retrieval can probably be optimized using a sliding window approach.
• Investigate whether _calc_match_score can be harmonized between de novo and DB modes.