Add LLM memory quality eval gate and promote OpenAI-agent readiness updates

refactor: extract startup summary lines into runtime helper

Extract startup schema and blocker helpers from main runtime