improve evaluate outputs

[wsnoble]

Currently, the evaluate option simply adds three lines to the log file, like this:

2025-04-11 17:32:38,752 INFO [casanovo/MainProcess] model_runner.log_metrics : Peptide Precision: 49.23% 2025-04-11 17:32:38,753 INFO [casanovo/MainProcess] model_runner.log_metrics : Amino Acid Precision: 71.18% 2025-04-11 17:32:38,753 INFO [casanovo/MainProcess] model_runner.log_metrics : Amino Acid Recall: 71.28%

I think we should also report average precision, and we should output a separate TSV file that allows for plotting of the precision-coverage curve. This could include the Casanovo score, ground truth and predicted peptides, correct/incorrect label, and the precision.

[bittremieux]

I'd maybe add a few extra columns to the mzTab in that case, rather than exporting a second file. That way, all of the information is present in a single file. In your description for example the amino acid scores are not included, which are needed for an AA-level precision-coverage curve. And similarly there might be other analyses you could do for which some other data is needed, so I'd just keep everything in a single file.

[wsnoble]

Yes, I agree that putting additional columns (ground truth peptide, correct/incorrect, precision) is a good idea. If we do that, then we probably want to sort the output by score.