Feat: support OOD dual label

[anyangml]

This is a follow up PR of #372 . For OOD-FF datasets, we provide both PBE level and wB97MV level labels as separate datasets. We allow model to select the corresponding data path based on the model branch.

[Copilot]

Pull Request Overview

This PR refactors the model instantiation workflow to support domain-specific model heads (molecules vs. materials) for models trained with OMol25 datasets. The key changes include delaying model instantiation until the specific task domain is known, enabling proper selection of model heads and DFT labels.

• Refactored model gathering to separate parameter collection from model instantiation
• Added support for domain-specific model heads (model_domain attribute) for molecules and materials
• Modified the calc property from cached_property to regular property with lazy initialization and setter support

Reviewed Changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
lambench/workflow/entrypoint.py	Split gather_models() into gather_model_params() and gather_model() to enable delayed model instantiation based on task domain
lambench/models/basemodel.py	Added supports_omol and model_domain attributes to support domain-specific model configurations
lambench/models/ase_models.py	Changed calc from cached_property to property with setter, added domain-based head selection for UMA and DP calculators, and logic to select appropriate DFT labels
lambench/tasks/base_task.py	Modified test_data type to support both single Path and dict of Paths for multiple DFT labels
lambench/tasks/direct/direct_tasks.yml	Added AQM task with PBE and wB97 test data paths, moved ANI and MD22 to deprecated section
lambench/metrics/utils.py	Updated to use new gather_model_params() and gather_model() functions
tests/workflow/test_entrypoint.py	Updated test fixture to return dict instead of DPModel instance, aligned with new parameter-based approach
tests/tasks/calculator/test_nve_md.py	Simplified test fixtures by removing separate calculator fixture and accessing calculator through model

Comments suppressed due to low confidence (1)

lambench/models/ase_models.py:37

• The documentation still describes calc as a 'cached property', but it has been changed to a regular property with lazy initialization. Update the documentation to reflect this change.

        calc (Calculator): A cached property that initializes and returns an ASE Calculator

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov]

Codecov Report

❌ Patch coverage is 14.28571% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 65.59%. Comparing base (04e9340) to head (fb1ada5).
⚠️ Report is 10 commits behind head on main.

Files with missing lines	Patch %	Lines
lambench/models/ase_models.py	0.00%	6 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #373   +/-   ##
=======================================
  Coverage   65.59%   65.59%           
=======================================
  Files          35       35           
  Lines        1613     1613           
  Branches      195      195           
=======================================
  Hits         1058     1058           
  Misses        511      511           
  Partials       44       44           

☔ 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.