RAG Evaluations MVP

Description

LFAI needs a framework for evaluations in order to:

The MVP for this evaluations framework will be limited in scope to the following evaluation scenarios:

Chat - regular chat-based interactions with the llm without added context
Chat with your docs - RAG-enabled chat interactions that require retrieved context

This framework can be expanded upon as LFAI develops, but this epic covers an MVP that accomplishes the above goals.

[x] Decision of what toolset should be used to handle evaluations (Deepeval)
- https://github.com/defenseunicorns/leapfrogai/issues/196
[x] Identification of 1 or more existing evaluation datasets for rapid prototyping
- https://github.com/defenseunicorns/leapfrogai/issues/715
[x] Creation of a custom dataset that is mission-focused while remaining openly available for more targeted evaluations
- https://github.com/defenseunicorns/leapfrogai/issues/716
[x] Determine a list of open-source models to evaluate on
- https://github.com/defenseunicorns/leapfrogai/issues/717
[x] Establishment of which evaluation metrics and custom heuristics should be used
- https://github.com/defenseunicorns/leapfrogai/issues/718
[x] Decision of what evaluator llm should be used for llm-as-a-judge evaluations
- https://github.com/defenseunicorns/leapfrogai/issues/719
[x] Decision on how evals will be run (workflow? ad-hoc python script? notebook?) and where data will be stored
- https://github.com/defenseunicorns/leapfrogai/issues/720
[x] Determine what the "Unicorn model card" should look like (i.e how the evaluation results are delivered/conveyed)
- https://github.com/defenseunicorns/leapfrogai/issues/721
[x] ADR for LFAI Evals Framework
- https://github.com/defenseunicorns/leapfrogai/issues/823
[x] Add Needle in a Haystack Evaluations
- https://github.com/defenseunicorns/leapfrogai/issues/944
[x] Add LLM-as-judge Evaluations
- https://github.com/defenseunicorns/leapfrogai/issues/957