[Torchax] Add initial support for loading mxfp4

[kyuyeunk]

Description

Add support for loading mxfp4 weights in torchax path but dequantize into bf16 for now.

Verified that it returns correct numeric outputs with openai/gpt-oss-120b

Also verified that performance is the same with bf16 version unsloth/gpt-oss-120b-BF16

Referenced mxfp4 pr in JAX path for some of functionalities: #992

Tests

pytest -v -s tests/layers/vllm/test_mxfp4.py

MODEL_IMPL_TYPE=vllm examples/offline_inference.py --task=generate --model=openai/gpt-oss-120b --max-model-len=1024 --max-num-batched-tokens=1024 --max-num-seqs=128 --no-enable-prefix-caching --tensor-parallel-size=4 --gpu-memory-utilization=0.8

https://buildkite.com/tpu-commons/tpu-inference-ci/builds/5302

Checklist

Before submitting this PR, please make sure:

• I have performed a self-review of my code.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have made or will make corresponding changes to any relevant documentation.

[github-actions]

Description

Start with a short description of what the PR does and how this is a change from
the past.

The rest of the description includes relevant details and context, examples:

• why is this change being made,
• the problem being solved and any relevant context,
• why this is a good solution,
• some information about the specific implementation,
• shortcomings of the solution and possible future improvements.

If the change fixes a bug or a Github issue, please include a link, e.g.,:
FIXES: b/123456
FIXES: #123456

Tests

Please describe how you tested this change, and include any instructions and/or
commands to reproduce.

Checklist

Before submitting this PR, please make sure:

• I have performed a self-review of my code.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have made or will make corresponding changes to any relevant documentation.

[kyuyeunk]

cc @amishacorns @bzgoogle