feat: ship llm-only turn understanding and memory governance upgrades

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

Add operator console, observability, skills, and resilience/security hardening

Add retention windows and startup credential safety warnings

Add HA todo/timer/area tools plus reminders and calendar helpers

Harden HA conversation policy/taxonomy and persist timers

Harden MemoryStore limit parsing for direct callers

Make task plan writes transactional and add rollback fault-injection tests

Add deep roadmap and harden summary/action history/memory lock behavior

Improve persona style lookup and isolate test memory state

Harden memory state transitions and maintenance hooks

Add detailed roadmap and implement Wave A hardening

Harden memory decoding and audit serialization

Harden memory and tool argument handling