Add vllm bench [latency, throughput] CLI commands (#16508)

Signed-off-by: mgoin <mgoin64@gmail.com>

tests/benchmarks/test_throughput_cli.py

@@ -0,0 +1,19 @@
+# SPDX-License-Identifier: Apache-2.0
+import subprocess
+
+import pytest
+
+MODEL_NAME = "meta-llama/Llama-3.2-1B-Instruct"
+
+
+@pytest.mark.benchmark
+def test_bench_throughput():
+    command = [
+        "vllm", "bench", "throughput", "--model", MODEL_NAME, "--input-len",
+        "32", "--output-len", "1", "--enforce-eager", "--load-format", "dummy"
+    ]
+    result = subprocess.run(command, capture_output=True, text=True)
+    print(result.stdout)
+    print(result.stderr)
+
+    assert result.returncode == 0, f"Benchmark failed: {result.stderr}"