Add evalutation metrics and reorganize training hooks

[melihyilmaz]

Changes in this PR:

• Added AA and peptide level precision and recall metrics calculation throughevaluate.py, also logging the new metrics during validation.
• Added de novo sequencing during validation
• Reorganized Lightning hooks for on_train_epoch_end() and on_validation_epoch_end() so that we can now run validation during training. The downside is we have to provide a validation loader during training.
• Added option to provide tensorboard.SummaryWriter object to the model to more easily organize training curves.
• As a temporary solution to 1) peptide predictions with multiple stop tokens, i.e. for which decoding apparently doesn't stop at the first one 2) peptide predictions of length 0, filter out those spectra before eval. metric calculation

[melihyilmaz]

Changes in this PR:

• Added AA and peptide level precision and recall metrics calculation throughevaluate.py, also logging the new metrics during validation.
• Added de novo sequencing during validation
• Reorganized Lightning hooks for on_train_epoch_end() and on_validation_epoch_end() so that we can now run validation during training. The downside is we have to provide a validation loader during training.
• Added option to provide tensorboard.SummaryWriter object to the model to more easily organize training curves.
• As a temporary solution to 1) peptide predictions with multiple stop tokens, i.e. for which decoding apparently doesn't stop at the first one 2) peptide predictions of length 0, filter out those spectra before eval. metric calculation
• Added spectrum_utils to setup.cfg