[Compressed Tensors] Add XPU wNa16 support

[yiliu30]

Purpose

cd vllm
python examples/offline_inference/basic/generate.py  \
    --model Intel/Qwen3-8B-W4A16-G128-AutoRound-LLMC-TEST-ONLY \
    --gpu_memory_utilization 0.75 \
    --enforce-eager 

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.

[gemini-code-assist]

Code Review

This pull request introduces support for wNa16 compressed tensors on XPU by adding a new IPEXwNa16LinearKernel. The changes are mostly self-contained in a new file and registration of the new kernel. However, I've found a critical issue in the implementation of the new kernel within vllm/model_executor/layers/quantization/kernels/mixed_precision/ipex.py. The input feature size for the underlying IPEX linear layer is calculated incorrectly, which will likely lead to runtime failures or incorrect computations. I have also pointed out a confusing and redundant variable assignment that should be cleaned up. Please see the detailed comments for suggestions on how to fix these issues.

[hshen14]

@robertgshaw2-redhat @mgoin Please help review. The PR is to support the quantized models with compressed tensor format (e.g., quantized by LLM-C/AutoRound) for Intel GPUs.

[xuechendi]

@mgoin @robertgshaw2-redhat , may you help to take a look?

[jikunshang]

LGTM for xpu part.

[mgoin]

Great work! Sorry for missing this