Add ragged all-to-all pipeline correctness coverage

[huytransformer]

Description

Adds pipeline-parallel correctness coverage for DeepSeek dropless expert parallelism, which exercises the ragged_all_to_all path.

This also fixes the pipeline test helper so single_pipeline_stage_class is preserved. Previously the helper constructed the requested stage class, then unconditionally replaced it with SimpleDecoderLayerToLinen, so custom layer tests did not actually run the requested layer.

FIXES: b/430135480

Tests

• python3 -m py_compile tests/integration/pipeline_parallelism_test.py
• git diff --check

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run end-to-end tests tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!