Add trtllm_mla and cutlass_mla for ragged fmha for chunked prefill (#9480)

python/sglang/srt/models/deepseek_v2.py

@@ -999,6 +999,8 @@ class DeepseekV2AttentionMLA(nn.Module):
             attention_backend == "flashinfer"
             or attention_backend == "fa3"
             or attention_backend == "flashmla"
+            or attention_backend == "trtllm_mla"
+            or attention_backend == "cutlass_mla"
         ):
             # Use MHA with chunked KV cache when prefilling on long sequences.
             sum_extend_prefix_lens = (