Rework rank score file reading

[dnil]

Rank score files are currently read as MIP rank score model files with version number in the file name from the references-files repo. This will need generalisation. One option would be to read the rank score files as part of the upload, including the version in them.

Cf eg #3777 for other rank score files, but e.g. nf-core/raredisease and balsamic will also soonish have their own. And tending to pipelines with multiple genome versions without changing the model may not be possible (e.g. different population databases etc).

The exact implementation is up for discussion, but opening for passing the rank files directly with full URLs (and optionally as local file names) in the case load config seems straightforward. The current with version, where base URLs are added should be preserved as a fallback for old cases.