Qwen4-27B PR found

PR: github.com/analytics-zoo/vllm-xpu-kernels/pull/22 The PR author seems to have contributed to intel repos a lot and VLLM. He also shares his name (might be the same person?) with a person who is either a CTO somewhere at alibaba or the Director of Zhejiang Lab. He has worked on Qwen3.8 before this. The PR also implies the KV cache size for Qwen4-27B is not decreasing.