Extend 'create_judge' to handle chat evaluations (prompt is a list)

[rchan26]

For some models, you are able to pass in a list to the prompt field where we sequentially input the items in the list as different messages in a chat interface. The response field is therefore also a list.

Currently, the create_judge command only works for single prompt and response. We can implement a simple chat evaluation by presenting the judge LLM the conversation.