support for multi-turn qa dataset

[ashishkamra]

Hi Team,
I was wondering if we can add support for the following multi-turn datasets

1. https://github.com/LMCache/LMCache/tree/dev/benchmarks/multi-round-qa
2. https://github.com/ekwinox117/multi-challenge
3. https://huggingface.co/datasets/bryanlincoln/infoquest

so that we don't have to run their benchmarking scripts for these specific datasets that are becoming quite important to evaluate cost/accuracy tradeoffs in agentic workflows.

cc @sjmonson @dagrayvid @markurtz