Modular dependency management: optional extras for expanded modules

[enriquea]

Context

As hvantk expands with new modules (HGC, ancestry, annotation pipelines, etc.), some dependencies bring heavy transitive requirements that can fail on hosts without system-level dev packages. For example, gnomad pulls in ga4gh.vrs[extras] → hgvs → psycopg2, which requires gcc, libpq-devel, and python-devel to compile from source — a pain for collaborators on different OS environments.

Current state

• gnomad has been moved to an optional extra under hgc (the only actual import is already guarded with try/except ImportError in hvantk/hgc/converters.py)
• Base poetry install now works without system-level C/PostgreSQL dev libraries

Going forward

As the tool grows, modules should follow this pattern:

1. Core dependencies (hail, click, pandas, etc.) stay required — they are needed by all users
2. Module-specific heavy dependencies should be declared as optional extras in pyproject.toml
3. Code imports of optional packages must be guarded with try/except ImportError and clear error messages guiding the user to install the relevant extra
4. Install instructions in README should document available extras, e.g.:

   poetry install                    # core functionality
   poetry install --extras hgc       # HGC pipeline (includes gnomad, plotting)
   poetry install --extras viz       # visualization support

TODO

• Review current extras (viz, interactive, hgc, psroc) and ensure they cover all optional modules
• Audit all imports for optional packages and add guards where missing
• Update README install section with extras documentation
• Consider whether scikit-learn should also be optional (only needed by specific modules)
• Test clean install on a minimal environment (no system dev packages) to verify base install works