Separate LLM and RAG evaluation

Fix database function structure

Fix evaluation database mapping

Add error traceback to debug 500

Fix: update aggregator with RAG evaluators

Add RAG evaluators - context relevance and recall + updated main

Fix: short factual answers no longer wrongly flagged as Irrelevant

Auto-create data folder for SQLite on startup

Fix Dockerfile name, remove pycache

Initial deploy of LLM eval API