Add auto-annotation mechanism!

[akshaylive]

LLM-as-a-Judge ideally needs to be aligned with humans. Create a feature to capture human feedback using some sort of templating mechanism. It should also have configurable DB to store and also a way to tag metrics and sub-metrics.