Refactor layer-wise tests with torch

[hengtaoguo]

Description

Reverts the standalone cpu_torch_reference_tests workflow and integrates gemma4_layers_test.py into the standard cpu-unit tests. The test is configured to run only on scheduled CI runs.

• Removed maxtext_cpu_torch_reference_tests workflow.
• Enabled gemma4_layers_test.py in pytest.ini, and marked it as cpu_only and scheduled_only.
• Configured torch installation for cpu-unit only during scheduled runs.

TODO: After this atomic PR, follow up with another PR refactoring import torch for other layer-wise tests, and add them to scheduled runs cleanly.

Tests

JAX_PLATFORMS=cpu python -m pytest tests/unit/gemma4_layers_test.py

Note: GitHub Actions Scan is failing due to a very recent Google-wise security rollout on 07-27-2026.

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.

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.