Support FlashMLA backend cuda graph (#4514)

Co-authored-by: yinfan98 <1106310035@qq.com>
Co-authored-by: Hongbosherlock <hongbosherlock@gmail.com>
Co-authored-by: ispobock <ispobaoke@163.com>

python/sglang/srt/layers/attention/utils.py

@@ -49,7 +49,6 @@ def create_flashmla_kv_indices_triton(
     kv_indices_ptr,
     req_to_token_ptr_stride: tl.constexpr,
     kv_indices_ptr_stride: tl.constexpr,
-    max_pagesize: tl.constexpr,
 ):
     PAGED_SIZE: tl.constexpr = 64
     BLOCK_SIZE: tl.constexpr = 4096