feat: register protein sequence consequence/canonical allele catvars

[jsstevenson]

close #453

Decisions

• /categorical_variants not /categorical_variations. I don't even know which is correct at this point
• the "reference" to the external resource is just the ID of the catvar itself. So rather than making a catvar with an "xref" to a clingen allele record, the id field is just clingen.allele:CA12345. And similarly, fetching that specific catvar is GET /categorical_variants/canonical_alleles/clingen.allele:CA12345
• Rather than a general "drop in a catvar, have some indication of its type, let anyvar figure it out" endpoint, I just made resource type-specific registration endpoints, I think that's better for our sanity and avoids inventing any new conventions. Similarly, there are separate sql tables for each categorical variant type. I had wanted to be slick but it probably makes more sense to just use a new table for each type.
• Make new classes for PSQ's/CA's instead of reusing ga4gh.cat_vrs recipes, a couple reasons
  1. we want to be stricter about some validation things (ID is required, len(constraints) == 1)
  2. we might want to interpret relations differently than the recipes do, now or in the future
• Validating molecule type is done outside of the pydantic field_validator because it might require seqrepo access. Alternatives here could be 1) requiring the submitter to explicitly declare molecule type and trusting it, or 2) injecting validation context into the fastapi route validation logic, but that looked like it was going to be insane. It's also possible that this function will duplicate work from the projection end point so whichever one gets merged second can resolve that.
• Request paths below -- also add basic GET by ID for round-trip testing and also because it just seems reasonable and enabled roundtrip testing
• Haven't added any relations to the input or output objects yet, pending final determination on how we implement our search functions. But broadly, I think you could just make sure to add in the right values when the object is coming back out to the user. Not sure if it's worth validating them on the way in.
• Haven't updated docs yet, pending final search endpoint behavior
• Didn't add a Snowflake implementation because a) scary and b) maybe we are ending snowflake support?

Routes

PUT /categorical_variants/canonical_alleles -- to register a CA
GET /categorical_variants/canonical_alleles/<id> -- to fetch a CA by ID
PUT /categorical_variants/protein_sequence_consequences -- to register a PSQ
GET /categorical_variants/protein_sequence_consequences/<id> -- to fetch a PSQ by ID

[jsstevenson]

pulled in latest commits to main + added more tests

[theferrit32]

Can we add a moleculeType="protein" or moleculeType=vrs.MoleculeType.PROTEIN here?

[jsstevenson]

Yeah. I meant to do this before, but also updated the molecule type validation function to trust moleculeType blindly.

[theferrit32]

We could add more around this if we do #478