[Feature] optimize group gemm

[zhyncs]

Checklist

• 1. If the issue you raised is not a feature but a question, please raise a discussion at https://github.com/sgl-project/sglang/discussions/new/choose Otherwise, it will be closed.
• 2. Please use English, otherwise it will be closed.

Motivation

Rewrite the Grouped GEMM used by LoRA with cuBLAS 12.5 in sgl-kernel for improved speed.

https://developer.nvidia.com/blog/introducing-grouped-gemm-apis-in-cublas-and-more-performance-updates/
https://github.com/zhihu/ZhiLight/blob/main/src/nn/linear/gemm_grouped.cpp

Related resources

No response

[Fridge003]

Thanks, this seems to be a good idea!