Full implementation of Casanovo-DB

[VarunAnanth2003]

Addresses all previous comments in PR #325.

High-level changes:

• Removed all unnecessary subclassing
• Integrated methods into old Casanovo code as much as possible
• Removed unnecessary flags for db-search
• Moved PSM batching into dataloader as opposed to the model
• Made .mztab output for Casanovo-DB consistent with specifications
• Added tests for all new functionality

Things to Note:

• Works with a set list of modifications (all Casanovo mods), cannot be modified at the moment
• Have not filled out unique, database, and database_version columns of the output .mztab for db-search mode

[bittremieux]

Nice work. It's starting to look quite nice.

Several smaller comments that will be easy to address.

One major comment I have is that we should try to abstract the database functionality a bit more. Now a list of tuples with peptide, mass, protein is passed through several levels in the code. Instead, this can be encapsulated by a ProteinDatabase class with functions cleave, get_candidates, etc. (examples) that provides these functionalities in a much more coherent and understandable fashion.

After that initial implementation, you could also simplify some parts of the database functionality. For example, passing the protein identifier through all inference steps is unnecessary. Instead, the ProteinDatabase can retain this mapping internally, only the peptides are used during inference, and upon exporting this mapping is used to retrieve the relevant protein information.

I suggest starting with drawing a schematic of how this class would work, where it receives information from and where it will be called, and the data flow among those steps. And only after you have this clear mental overview, start with the implementation.

[VarunAnanth2003]

@bittremieux All comments you left should be addressed in the most recent commit