refactor unet_spatiotemporal tests

[akshan-main]

What does this PR do?

Part of the ongoing modeling-test migration (following #13369 and #13153). Migrates the UNetSpatioTemporalConditionModel test suite to the mixin-based structure (Config + ModelTesterMixin / TrainingTesterMixin / MemoryTesterMixin / AttentionTesterMixin). First of the UNet test migrations.

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? (Discussed on Slack with @sayakpaul @DN6)
• 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?

@sayakpaul @DN6

[akshan-main]

The model's _repeated_blocks is empty, so the compile mixin is not included. A few UNetTesterMixin-era tests have no equivalent in the new mixins and were dropped: test_xformers_enable_works (xformers is deprecated), test_from_save_pretrained_dynamo, test_deprecated_kwargs, and the set_attn_processor determinism tests. The model-specific tests (num_attention_heads/cross_attention_dim tuple configs, pickle, gradient-checkpointing set) are kept. Happy to keep any of the dropped ones as overrides if you'd prefer.

[sayakpaul]

Thanks for the test refactor. I think it's fine to drop those tests since these models aren't used that much anyway.

[sayakpaul]

Ran the tests locally as well and we're good!