Fix Ulysses SP backward with SDPA

[zhtmike]

What does this PR do?

Solve the issue #13319.

There are two bugs:

• grad_out is already in BSHD format, which matches the shape of out. Therefore, we do not need to permute grad_out again.
• autograd.Function.backward() does not store the gradient by default, so we need to enable it manually.

After the fix, running torchrun --nproc-per-node 2 toy_train.py --enable-sp can get the expected result

loss=1.351188

And we add the test coverage for backward ops with context parallel.

Tested with TestQwenImageTransformerContextParallel, TestFluxTransformerContextParallel and TestFlux2TransformerContextParallel.

Fixes # (issue)

• Fix Ulysses SP backward with SDPA

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?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

@sayakpaul

[sayakpaul]

Ran the tests and they pass as well. Great work!

[zhtmike]

Hi @sayakpaul, do I have anything else to help for this PR?

[sayakpaul]

Gonna let the CI run and merge afterward. Sorry for the delay!

[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.