Support VL Model Benchmarking

[rgreenberg1]

Description:
Add support into GuideLLM to benchmark VL Models (Images in, Text out).

Acceptance Critieria

[huijiao1120]

This would be a great new feature to add, also want to ask if we can support benchmarking with multiple images per request, not necessarily just one image.