Predictors cleanup

[stephaniesamm]

• Every pipeline runs with uv run
• Shared biological resources are downloaded on the first run that needs them.
• Shared resources are reused by later pipelines and are ignored by Git.
• Predictor-specific downloads and intermediate files stay inside each predictor’s data/ directory and are ignored.
• Standardized prediction tsvs stay under data/predictions// and remain ignored.
• Pipeline logs also go under data/predictions//, but logs are tracked by Git.
• Biomart mapping stays under mirDB.
• Harmonised CLI commands
• Updated all relevant READMEs
• Added data/common_resources/README.md that explains the reason of existence of this directory.
• .gitignore cleanup
• Reran all predictors

[zacharopoulou]

Important question: Is the check of "new"/"old" predictions be done? Are we ok?

[stephaniesamm]

@zacharopoulou if all looks good, we can merge and I can publish the new Zenodo version: https://zenodo.org/uploads/21671476