feat(engine): support Qwen3-VL with native AWEX colocate

[sitabulaixizawaluduo]

Description

This PR adds AWEX colocate support for Qwen3-VL Dense and MoE models.
Key Changes:

• Passes the complete multimodal Hugging Face configuration to AWEX, including both text_config and vision_config.
• Reads router_dtype from the nested text configuration used by Qwen3-VL.
• Delegates nested configuration parsing and weight conversion to AWEX’s native Qwen3-VL implementation.
• Adds unit tests covering:
  • Dense and MoE model registration.
  • Composite configuration propagation.
  • Vision tower TP sharding.
  • Architecture fallback and router dtype handling.

This PR only contains Qwen3-VL colocate integration.

AWEX has not yet released an official version supporting the Qwen3-VL model. To use it, run:
uv pip install --reinstall --no-deps "awex @ git+https://github.com/inclusionAI/Awex.git@5c6e802e53919818e979f3e03a6c02b1abeb89da"

N/A

Type of Change

• 🐛 Bug fix
• ✨ New feature
• 💥 Breaking change
• 📝 Documentation update
• ♻️ Refactoring
• ⚡ Performance improvement
• ✅ Test coverage improvement

Checklist

• I have read the
  Contributing Guide
• Pre-commit hooks pass (pre-commit run --all-files)
• Relevant tests pass; new tests added for new functionality
• Documentation updated (if applicable; built with ./docs/build_all.sh)
• Branch is up to date with main
• Self-reviewed via /review-pr command
• This PR was created by a coding agent via /create-pr
• This PR is a breaking change

Breaking Change Details (if applicable):

Additional Context

---

Need help? Check the
Contributing Guide
or ask in GitHub Discussions!

[Le8r0nJames]

Once an AWEX version with Qwen3-VL support is released, please update the dependency, replace pytest.importorskip("awex.models.qwen3_vl") with a regular import, and rerun the tests against the released package.

[Le8r0nJames]

LGTM