Reduce production DB connection burst during deploys

[ewlarson]

Summary

• Set production DB_POOL_MIN=0 and DB_POOL_MAX=1 to avoid eager per-worker DB connections during web container startup.
• Reduce production FastAPI workers from 4 public + 6 internal to 2 public + 2 internal.
• Add a short comment explaining the rolling deploy connection-headroom constraint.

Why

The 0.8.9 deployment logs show FastAPI workers failing at startup with:

asyncpg.exceptions.TooManyConnectionsError: sorry, too many clients already

The production web container currently starts 10 FastAPI workers, and during a rolling deploy old and new containers overlap. Each worker creates its own DB pool, so startup can exceed Postgres connection headroom before the new container becomes healthy.

Validation

• git diff --check
• Verified asyncpg permits min_size=0 for pools.

No backend tests were run for this config-only change.