feat: add GP-age epigenetic clock

[marcbal77]

Summary

• Add GP-age clock (Gaussian Process regression-based age prediction)
• 6 model variants: GPAge10, GPAge30, GPAge71, GPAgeA, GPAgeB, GPAgeC
• GPy added as optional dependency (pip install biolearn[gpage])

Reference

• Paper: https://www.cell.com/cell-reports-methods/fulltext/S2667-2375(23)00211-4
• GitHub: https://github.com/mirivar/GP-age
• License: MIT (confirmed with authors)

[marcbal77]

Note: The .json.zip extension is misleading - they're actually gzip-compressed JSON files. GPy handles this natively. I noticed this is different than the linear clocks managed as csv's.

[sarudak]

I'm not clear what this needs to be ready to merge

[marcbal77]

I ran into some error locally and I could not repeat it and I looked further into GPy, I just recall it had to do with the .json.zip extension and could not replicate. Let me rebase and make test/format and then take out of draft state.

[marcbal77]

1. Initially observed an error when trying to open the model file directly - something related to gzip decompression (couldn't replicate the exact error)
2. Checked file types - confirmed .json.zip files are actually gzip-compressed (not zip):
   $ file biolearn/data/GP-age_model_10_cpgs.json.zip
   gzip compressed data, max compression
3. Tested model loading:
   from biolearn.model_gallery import ModelGallery
   gallery = ModelGallery()
   model = gallery.get('GPAge10', imputation_method='none')

Loads successfully, 10 CpG sites

4. Ran full test suite: 161 passed, 5 skipped

I could not reproduce the original gzip error. GPy's load_model() handles gzip natively despite the misleading .json.zip extension. Tests pass, models load correctly. Rebased on master and marked ready for review.

[sarudak]

I think this is ready to merge. Would like to discuss the separation of gpy as an optional dependency.

[marcbal77]

We concluded to leave GPy as a dependency due to compatibility issues and dependencies on other large libraries.