Enable vLLM scheduler specific Metrics

[rgreenberg1]

Description: The purpose of this ticket is to surface vLLM specific scheduler metrics to provide users with a full array of data on the vLLM service running. 

User Story:
As a ml developer, I want to be able to see the vLLM specific scheduler metrics so that I can ensure that my benchmarks are taking full advantage of what vLLM has to offer and that my service is running optimally before I move to production. 

Acceptance Criteria: 

• Stream the metrics from Prometheus to the output format script: 
• Queue size
• KV cache usage
• Requests in Waiting, Swap, Running
• Create a new CLI output table dedicated for the vLLM specific metrics: 
• • needs further scoping