Use folder for each task

[dhruvdcoder]

Summary

Task code under xlm.tasks is reorganized so each task has its own package directory (src/xlm/tasks/<name>/__init__.py). Public Hydra dotted paths stay the same (for example xlm.tasks.owt.preprocess_fn, xlm.tasks.safe_molgen.DeNovoEval), so existing YAML configs under src/xlm/configs/lightning_train/ do not need bulk updates.

This PR also adds contributor-facing documentation for wiring new tasks and fixes a few stale molgen references in docs and examples.

Details

• Layout: Former flat modules (cfg.py, owt.py, …) are renamed into subpackages with __init__.py. safe_molgen now contains the lazy facade, _safe_molgen_impl.py, and zinc_len.pkl; the lazy import targets xlm.tasks.safe_molgen._safe_molgen_impl.
• xlm.tasks package marker: src/xlm/tasks/__init__.py documents that this file is not a barrel export of every task symbol.
• Packaging: setup.py package_data path for the pickle is corrected to tasks/safe_molgen/zinc_len.pkl (relative to the xlm package).
• Configs: xlm-models/mlm/configs/experiment/safe_mlm.yaml length_ref_file updated to src/xlm/tasks/safe_molgen/zinc_len.pkl.
• Docs: New docs/guide/adding-a-task.md and a link from docs/index.md.
• Wiki: wiki/eval.md examples updated from xlm.tasks.molgen / “molgen” wording to safe_molgen / SAFE molecular generation where appropriate.
• Small fix: composite_eval docstring example now references xlm.tasks.safe_molgen.DeNovoEval.

Backward compatibility

• Python / Hydra: import xlm.tasks.<name> and _target_: xlm.tasks.<name>.<callable> behavior is preserved for unchanged symbol names.
• Filesystem: Anything that referenced src/xlm/tasks/zinc_len.pkl should use src/xlm/tasks/safe_molgen/zinc_len.pkl (updated in-repo for safe_mlm).

Verification

• Import smoke for task subpackages and ZINC_LENGTH_REF_FILE resolution.
• hydra.utils.get_method("xlm.tasks.owt.preprocess_fn") succeeds.
• Targeted pytest (for example datamodule / utils tests) as run locally before opening the PR.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 0% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 33.49%. Comparing base (94fb33c) to head (2a95c6e).

Files with missing lines	Patch %	Lines
src/xlm/tasks/safe_molgen/__init__.py	0.00%	1 Missing ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #41      +/-   ##
==========================================
- Coverage   34.27%   33.49%   -0.79%     
==========================================
  Files         120      120              
  Lines       10698    10698              
  Branches     1400     1400              
==========================================
- Hits         3667     3583      -84     
- Misses       6890     6976      +86     
+ Partials      141      139       -2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.