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

test/manual/dsv4/test_h200_fp8_flash.py

@@ -53,7 +53,7 @@ class TestH200Fp8FlashBalanced(DSV4FlashAime25TestBase):
         "1",
         "--speculative-num-draft-tokens",
         "2",
-        "--cuda-graph-max-bs",
+        "--cuda-graph-max-bs-decode",
         "128",
         "--max-running-requests",
         "128",
@@ -77,7 +77,7 @@ class TestH200Fp8FlashMaxThroughput(DSV4FlashAime25TestBase):
         "--enable-dp-attention",
         "--moe-a2a-backend",
         "deepep",
-        "--cuda-graph-max-bs",
+        "--cuda-graph-max-bs-decode",
         "128",
         "--max-running-requests",
         "256",