feat(l2g): implement new training strategy splitting between EFO/gene pairs and with cross validation

[ireneisdoomed]

✨ Context

The changes here reflect the training strategy used for training the latest L2G model (presented in the Genetics meeting on 21/11 and discussed here).

I've essentially changed the LocusToGeneTrainer.train function to follow that diagram with a 5 fold cross validation strategy set by default.

The most complicated part has been to integrate it with W&B Sweeps. Sweeps allow to group different runs together and are useful to test several parameter configurations and compare between them. They have a known issue where each cross validation fold keeps being overwritten, so tweaking this so that a Sweep shows results for each fold + the evaluation for the test set was the part that kept me busy. For the record, I took inspiration from here.

An example run can be found here: https://wandb.ai/open-targets/gentropy-locus-to-gene/sweeps/917iffr0?nw=nwuseropentargets

This closes opentargets/issues#3253

🛠 What does this PR implement

• New param in the L2G step: cross_validate. Whether to run this step. It is set to true by default. I've tested that the function works on both scenarios.
• Rewritten train: to implement the new strategy (described in docs). The key difference is the strategy to split between train and test sets, which takes into account gene/EFO pairs so that they are always kept separate.
• Changed hyperparameter_tuning to cross_validate. This function now not only accepts a grid of parameters to sweep over, but is the responsible for doing and tracking the cross validation. This is complex because I had to play a lot around the W&B config variables so that each sweep contained all my runs.
• Added best set of params to the dictionary of LocusToGeneStep config defaults and to the hyperparameters default dict inside the model.

🙈 Missing

🚦 Before submitting

• Do these changes cover one single feature (one change at a time)?
• Did you read the contributor guideline?
• Did you make sure to update the documentation with your changes?
• Did you make sure there is no commented out code in this PR?
• Did you follow conventional commits standards in PR title and commit messages?
• Did you make sure the branch is up-to-date with the dev branch?
• Did you write any new necessary tests?
• Did you make sure the changes pass local tests (make test)?
• Did you make sure the changes pass pre-commit rules (e.g poetry run pre-commit run --all-files)?

[project-defiant]

All looks good, although I am not 100% expert on the cross validation.