[FEAT][ROCm]: Support AITER MLA on V1 Engine (#17523)

Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com>
Co-authored-by: qli88 <qiang.li2@amd.com>
Co-authored-by: Hongxia Yang <62075498+hongxiayang@users.noreply.github.com>

vllm/engine/arg_utils.py

@@ -1319,6 +1319,7 @@ class EngineArgs:
             "FLASHMLA",
             "FLASHINFER",
             "FLASHINFER_VLLM_V1",
+            "ROCM_AITER_MLA",
         ]
         if (envs.is_set("VLLM_ATTENTION_BACKEND")
                 and envs.VLLM_ATTENTION_BACKEND not in V1_BACKENDS):