Support Hunyuan Dense V1 export

[hanbitmyths]

Summary

• Update the Hunyuan Dense V1 builder to bake Hunyuan dynamic NTK-alpha RoPE into rope_theta.
• Clear generic RoPE scaling metadata after baking the effective theta so the base builder does not apply another scaling path.
• Alias Hunyuan query/key layernorm modules to the existing q_norm/k_norm export path when needed.

Validation

• Ran py_compile for src/python/py/models/builders/hunyuan.py.
• Ran git diff --check for src/python/py/models/builders/hunyuan.py.

[Copilot]

Pull request overview

This PR updates the HunYuan Dense V1 model builder to better support export across Transformers config variants by baking Hunyuan dynamic NTK-alpha RoPE scaling into rope_theta and adapting Q/K norm module aliases.

Changes:

• Reads RoPE metadata from either rope_scaling or rope_parameters.
• Bakes Hunyuan alpha scaling into config.rope_theta and clears generic scaling metadata.
• Handles q_norm/k_norm aliases when actual modules are exposed as query_layernorm/key_layernorm.

[kunal-vaishnavi]

/azp run Integration Tests

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).