[tests] reorganize cache tests and clean memory between tests

[gante]

What does this PR do?

#37394 introduces a cache change, but there are several cache test issues at the moment, preventing a clean merge.

This PR:

1. Removes the @slow decorator from a few tests (change pulled from [Cache] Support compilable cache reuse with smaller batch sizes #37394)
2. Deletes test_static_cache_greedy_decoding_pad_right, as it doesn't make sense to generate with right padding (change pulled from [Cache] Support compilable cache reuse with smaller batch sizes #37394)
3. Moves torch.export() cache tests to its own class
4. Cleans device memory between test runs. On some devices, we were observing CPU weight offloading after a few tests, massively slowing down runtimes. Cleaning the device memory between tests fixes it.

---

Follow-up PR, before retesting #37394:

1. convert as many @slow tests into fast tests as possible
2. fix failing tests

[github-actions]

Hi 👋, thank you for opening this pull request! The pull request is converted to draft by default. The CI will be paused while the PR is in draft mode. When it is ready for review, please click the Ready for review button (at the bottom of the PR page). This will assign reviewers and trigger CI.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[zucchini-nlp]

Nice 🧼