Feature: build on the multi-chat playground to add a chatbot arena functionality to test 2 models side by side and rate the answers

[massi-ang]

This feature would help customers to evaluate different models using use case specific data and prompts and rate the answers.

This new functionality should also allow the user to setup the specific prompts to be used by the langchain RetrievalChain to better support the tuning of the system on specific customer data, tasks, etc.

Ref: https://chat.lmsys.org/