Create evaluation records from AI judges to database

Your AI evaluation outputs are scattered across sub-zap responses, leaving test cases untracked and slowing review. Store judge scores in a central table for fast review within minutes.