Add a script to run vLLM benchmark

[huydhn]

The script will pull vLLM repo, run its benchmarks, and upload the results to PyTorch ossci-benchmarks bucket to be ingested into PyTorch OSS benchmark database. On the devgpu, I have this one running in a while loop under screen:

#!/bin/bash

set -eux

while :
do
  HF_TOKEN=<REDACTED> ./run.sh main
  sleep 300
done

[huydhn]

@seemethere is OOTO this week, and as this code has already been running in the past few days. I'll merge it now and ask for forgiveness later so that I can add a workflow to benchmark on MI300x here.