fix(test): improve test_scan_layers_mismatch_tpu assertion

[mesakhcienet]

Description

Fixes the assertion in test_scan_layers_mismatch_tpu to expect the correct error message.

• Why is this change being made: The integration test test_scan_layers_mismatch_tpu was failing because the framework raises "Checkpoint does not match the model..." when attempting to restore a checkpoint with mismatched scan_layers in the pre-training workflow. The test was still asserting an outdated error message ("Configuration mismatch" or "Failed to restore checkpoint").
• The problem being solved: test_scan_layers_mismatch_tpu was producing false negatives (failing tests when the underlying code correctly caught the issue).
• Why this is a good solution: Instead of modifying the correct, working logic inside the pre-training workflow (which correctly bypasses metadata syncing for deterministic configuration matching and gracefully catches the mismatch during weight loading), we simply update the test's assertions to align with the framework's actual behavior.
• Specific implementation: Appended "Checkpoint does not match the model" to the list of accepted error strings in tests/integration/checkpointing_test.py::test_scan_layers_mismatch_tpu.
• Shortcomings/future improvements: We could potentially unify the error messages emitted by from_pretrained (metadata check) and train_main (weight alignment check) in the future to improve consistency.

Bug : https://b.corp.google.com/issues/532182112

Tests

• Testing instructions:
  Run the affected unit test using pytest to ensure it passes with the new assertions:

  pytest tests/integration/checkpointing_test.py::test_scan_layers_mismatch_tpu

  The test will successfully catch the ValueError and correctly validate the exception message.

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!