Add num_hidden_layers to t5gemma's top level config

[winnie0617]

What does this PR do?

This PR fixes a bug that for generating with cache using T5Gemma. Due to the nested structure of its config class, the field num_hidden_layers, which is used to initialize a dynamic cache, only exists under config.decoder.

This PR added a one line change to initialize config.num_hidden_layers using config.decoder.num_hidden_layers.

Fixes #41073

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,
  Pull Request section?
• 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?

@gante @Rocketknight1

Following up on this issue since I encountered the same error. Happy to provide more information or discuss alternative solutions to help get this merged :)

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: t5gemma

[ydshieh]

cc @gante ?

(but maybe we should handle this in the cache class instead of adding a new attribute only for this purpose ?)

[gante]

This PR is the actual fix 🤗

(this has been in review for a while, I'll ping a reviewer)