Expose LLM Evaluator Prompts

[joeferraratonic]

Currently the LLM evaluator prompts for a given metric are not exposed by the class that calculates the metric. Make it so that these prompts are explicitly accessible from the metric class.

This change will improve the tonic_validate integration in llama_index, as the evaluators in llama_index have methods that return these prompts. Once this change is made, a subsequent PR to llama_index should be made to add this functionality into the tonic_validate integration there.

Reference: run-llama/llama_index#10000 (comment)