Overhauled organization and interfaces

[wfondrie]

This PR makes sweeping changes across depthcharge, implementing some lessons learned using it with Casanovo and such. These include:

• Providing a proper tokenization class (also resolves #24 and #18)
• First-class support for ProForma peptide annotations, thanks to spectrum_utils and pyteomics.
• Moving preprocessing onto parsing instead of data loading (similar to @bittremieux's proposal in #31)
• Combining the SpectrumIndex and SpectrumDataset classes into one.
• Adding primitive dataclasses for peptides, peptide ions, mass spectra ... and even small molecules 🚀
• Adding type hints to everything and stricter linting with Ruff.
• Adding a ton of tests.
• Changing peak encodings. Instead of encoding the intensity using a linear projection and summing with the sinusoidal m/z encodings, now the intensity is also sinusoidally encoded and is combined with the sinusoidal m/z encodings using a linear layer.

• Tight integration with spectrum_utils 💪

  There may be things that I've missed, but this is a big PR that I'm excited about!

  Still TODO:

• [ ] Add real documentation.
• [ ] Clean up docstrings.
• [ ] Update release notes.

[codecov[bot]]

Codecov Report

Merging #34 (38cee4e) into main (2bfd7f1) will increase coverage by 5.82%. The diff coverage is 91.06%.

@@            Coverage Diff             @@
##             main      #34      +/-   ##
==========================================
+ Coverage   85.67%   91.50%   +5.82%     
==========================================
  Files          15       19       +4     
  Lines         705      965     +260     
==========================================
+ Hits          604      883     +279     
+ Misses        101       82      -19     

Impacted Files	Coverage Δ
depthcharge/tokenizers/peptides.py	68.65% <68.65%> (ø)
depthcharge/tokenizers/tokenizer.py	91.83% <91.83%> (ø)
depthcharge/data/spectrum_datasets.py	93.12% <93.12%> (ø)
depthcharge/transformers/peptides.py	94.54% <94.54%> (ø)
depthcharge/transformers/spectra.py	96.00% <96.00%> (ø)
depthcharge/primitives.py	97.31% <97.31%> (ø)
depthcharge/__init__.py	100.00% <100.00%> (ø)
depthcharge/constants.py	100.00% <100.00%> (ø)
depthcharge/data/__init__.py	100.00% <100.00%> (ø)
depthcharge/data/parsers.py	96.96% <100.00%> (+1.69%)	:arrow_up:
... and 8 more

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more

[wfondrie]

Yikes - peptide tokenization is much too slow right now to be useful. It takes about 36 ms on my Mac per peptide, which doesn't scale to thousands. 🤔

[wfondrie]

I'm going to go ahead and merge this. Please still leave a review and I'll fix problems in another PR!