Fix T5GemmaEncoder loading for transformers 5.x composite T5GemmaConfig

[DavidBert]

What does this PR do?

Fixes PRXPipeline.from_pretrained() failing with transformers 5.x due to T5GemmaConfig being refactored into a composite config.

In transformers 5.x, T5GemmaConfig became a composite config with encoder/decoder sub-configs (T5GemmaModuleConfig). When the PRX pipeline loads the text encoder, from_pretrained passes the composite T5GemmaConfig to T5GemmaEncoder.__init__, which expects flat attributes like attention_dropout that only exist on T5GemmaModuleConfig.

The fix wraps T5GemmaEncoder to extract config.encoder before passing it to the constructor.

Fixes #13142

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline?
• Did you read our philosophy doc (important for complex PRs)?
• Was this discussed/approved via a GitHub issue or the forum? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

@yiyixuxu @asomoza