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

test/manual/dsv4/test_dsv4_flash_sanity_dp4.py

@@ -65,7 +65,7 @@ class TestDSV4FlashTP4DP4(
                 "--enable-dp-attention",
                 "--moe-a2a-backend",
                 "deepep",
-                "--cuda-graph-max-bs",
+                "--cuda-graph-max-bs-decode",
                 "128",
                 "--max-running-requests",
                 "256",
@@ -101,7 +101,7 @@ class TestDSV4FlashTP4EP(
                 # No --enable-dp-attention by design: covers TP-attn path.
                 "--moe-a2a-backend",
                 "deepep",
-                "--cuda-graph-max-bs",
+                "--cuda-graph-max-bs-decode",
                 "128",
                 "--max-running-requests",
                 "64",
@@ -139,7 +139,7 @@ class TestDSV4FlashTP4DP4ChunkedPrefillLarge(
                 "deepep",
                 "--chunked-prefill-size",
                 "16384",
-                "--cuda-graph-max-bs",
+                "--cuda-graph-max-bs-decode",
                 "128",
                 "--max-running-requests",
                 "256",