DSV4 NNX support

[snehalv2002]

Description

This PR adds DeepSeek V4 (DEEPSEEK4) decoder block support and HyperHead routing integration to NNXDecoder.

Context & Implementation Details

• Instantiates DeepSeek4HyperHead on NNXDecoder when mhc_expansion_rate > 1 and decoder_block == DecoderBlockType.DEEPSEEK4.
• Routes the post-transformer hidden state reduction in NNXDecoder through self.hc_head(y) for DEEPSEEK4 instead of the default mhc_reduce(y).
• Added DecoderBlockType.DEEPSEEK4 to the list of supported decoder block types in NNXDecoder.

Tests

• Added unit test test_nnx_decoder_hyperhead_integration in tests/unit/deepseek_v4_vs_reference_test.py to verify NNXDecoder initializes hc_head as DeepSeek4HyperHead.
• Ran pytest tests/unit/deepseek_v4_vs_reference_test.py.

Checklist

• 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

❌ Patch coverage is 20.00000% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/maxtext/layers/nnx_decoders.py	20.00%	2 Missing and 2 partials ⚠️

📢 Thoughts on this report? Let us know!