Improved response time test and updated 03/06 for VLLM

[RHRolun]

closes #141

[guimou]

I will merge it as most things are already there, but we will need to modify some parameters for the Lab:

• Inference server URL
• Model name Both should be dynamic to allow the switch to central LLM at some point, and handle both standard serving and Single Stack Serving.