Adding support vLLM openai entrypoint benchmarking script

[Edwinhr716]

Currently the benchmarking script only works with the regular entrypoint of vllm. Since the benchmarking suite builds uses openai entrypoint instead, changed it to support that entrypoint
.

[vivianrwu]

For jetstream, can you change max_tokens: 1 to max_tokens: output_len under the benchmark_serving.py script? Other than that, lgtm for jetstream side changes!