[misc] Use --cuda-graph-max-bs-decode in tests, examples, and docs (#29591)

test/manual/ep/test_eplb_mori.py

@@ -69,7 +69,7 @@ common_args = [
     "12288",
     "--attention-backend",
     "aiter",
-    "--cuda-graph-max-bs",
+    "--cuda-graph-max-bs-decode",
     "32",
 ]
 

test/manual/ep/test_moe_deepep_eval_accuracy_large.py

@@ -32,7 +32,7 @@ class TestMoEDeepEPEvalAccuracyLarge(CustomTestCase):
                 "8",
                 "--moe-a2a-backend",
                 "deepep",
-                "--cuda-graph-max-bs",
+                "--cuda-graph-max-bs-decode",
                 "128",
             ],
         )

test/manual/ep/test_mooncake_expert_backup.py

@@ -58,7 +58,7 @@ class TestBackup(CustomTestCase):
                     "50",
                     "--chunked-prefill-size",
                     "512",
-                    "--cuda-graph-max-bs",
+                    "--cuda-graph-max-bs-decode",
                     "128",
                     "--max-running-requests",
                     "512",