Add human-inspectable to evaluate-llm.py output via `--show-annotations`

Fix issue with `&` character

Rename function to be more generic

Add evaluation