docs: note scale to zero disabled for cold start prevention (#424)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude <[e-mail]>

STATUS.md

···
 - increased `pool_size` from 5 → 10 (handle more concurrent cold start requests)
 - increased `max_overflow` from 0 → 5 (allow burst to 15 connections)
 - increased `connection_timeout` from 3s → 10s (wait for Neon wake-up)
+- disabled scale to zero on production compute (`suspend_timeout_seconds: -1`) to eliminate cold starts entirely
 
 **related**: this is a recurrence of the Nov 17 incident. that fix addressed the queue listener's asyncpg connection but not the SQLAlchemy pool connections.