Test MLPerf Inference Llama2-7b model on Windows and add to Github action

[arjunsuresh]

Since we have cleaned up the MLPerf inference pipeline and made most of it Windows compatible we should check and support Llama2-7b model using the reference implementation on Windows and once successful add a github action for it on Windows. Scheduling it twice a month is enough.

[arjunsuresh]

We now have this working on Windows but the run is too slow on the CPUs.