Add Flux2 DreamBooth prior preservation tests

[azolotenkov]

What does this PR do?

Adds DreamBooth LoRA smoke coverage for Flux2 and Flux2 Klein prior-preservation training with train_batch_size=2.

This covers the prior-preservation path where the effective batch is grouped as [instance samples..., class samples...]. The test helps guard against regressions in prompt/text embedding repetition and prior-loss splitting when the per-device train batch size is greater than 1.

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?

@sayakpaul

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Adds regression tests to ensure DreamBooth LoRA training for Flux2 (and Flux2 Klein) works with prior preservation when train_batch_size=2.

Changes:

• Added a new prior-preservation test case with train_batch_size=2 to test_dreambooth_lora_flux2.py.
• Added the equivalent prior-preservation test case to test_dreambooth_lora_flux2_klein.py.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
examples/dreambooth/test_dreambooth_lora_flux2.py	Adds a prior-preservation training test using batch size 2 and validates LoRA weights are produced.
examples/dreambooth/test_dreambooth_lora_flux2_klein.py	Adds the same prior-preservation batch size 2 test for the Flux2 Klein variant.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[azolotenkov]

I updated both tests to use a dedicated temporary class_data directory and copy two fixture images into it, so the tests no longer reuse the instance data directory for class images.