Fix: Scope unscale_gradients to relevant optimizers during clipping

[google-labs-jules]

Previously, clip_grad_norm_ and clip_grad_value_ would unscale gradients for all optimizers managed by the Accelerator, regardless of whether they were associated with the parameters being clipped. This was noted by a TODO comment in the code.

This change introduces a _get_associated_optimizers helper method that identifies which optimizer(s) manage the provided parameters. The clipping functions now use this helper to selectively unscale only the relevant optimizers. If no associated optimizer is found (e.g. for parameters not managed by any optimizer), it falls back to the previous behavior of unscaling all optimizers.

This improves efficiency and correctness in setups with multiple optimizers (e.g. GANs, actor-critic) where one might want to clip gradients for one network without affecting the accumulated gradients of another.

Verified with a reproduction script and existing tests.

---

PR created automatically by Jules for task 2963969511390292436 started by @iavinas

[google-labs-jules]

👋 Jules, reporting for duty! I'm here to lend a hand with this pull request.

When you start a review, I'll add a 👀 emoji to each comment to let you know I've read it. I'll focus on feedback directed at me and will do my best to stay out of conversations between you and other bots or reviewers to keep the noise down.

I'll push a commit with your requested changes shortly after. Please note there might be a delay between these steps, but rest assured I'm on the job!

For more direct control, you can switch me to Reactive Mode. When this mode is on, I will only act on comments where you specifically mention me with @jules. You can find this option in the Pull Request section of your global Jules UI settings. You can always switch back!

---

For security, I will only act on instructions from the user who triggered this task.

New to Jules? Learn more at jules.google/docs.