Restore Engram defaults in RLConfig

[csgoogle]

Summary

• add the existing Engram configuration mixin to RLConfig
• add a focused regression test for the disabled Engram defaults

Root cause

The dedicated RLConfig schema omitted Engram when RL configuration was split from MaxTextConfig. Shared decoder code reads engram_layers even when Engram is disabled, so RL models need those fields to satisfy the decoder configuration contract.

This change supplies the existing disabled defaults (engram_layers=[] and engram_max_ngram_size=3). It does not enable Engram or change Gemma 4 weights, architecture, or numerics.

Tests

• JAX_PLATFORMS=cpu pytest -q tests/post_training/unit/train_rl_test.py -k test_rl_config_includes_decoder_engram_defaults — passed
• git diff --check — passed

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.

[gemini-code-assist]

Caution

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

[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!