Replace hardcoded sleep with database readiness check in engine entrypoint

[KaranJagtiani]

The engine entrypoint script (deployment/engine/entrypoint.sh) uses a hardcoded sleep 30 to wait for the database before starting. This is unreliable — the database may not be ready after 30 seconds, or it may be ready much sooner, wasting startup time.

Scope

• Infra (deployment/engine/entrypoint.sh):
  • Replace sleep 30 with a loop that checks database connectivity before proceeding.
  • Use pg_isready (available in the container) or a lightweight Python check against the DATABASE_URL environment variable.
  • Add a configurable timeout (e.g., DB_WAIT_TIMEOUT, default 60 seconds) with a clear failure message if the database isn't reachable in time.
  • Add a similar readiness check for Redis using redis-cli ping or equivalent if the Redis client is available.

Acceptance criteria

• Engine waits for the database to be reachable before starting, rather than sleeping for a fixed duration.
• If the database is not reachable within the timeout, the script exits with a non-zero code and a clear error message.
• The timeout is configurable via an environment variable (DB_WAIT_TIMEOUT).
• Startup is faster when the database is already available (no unnecessary wait).
• No regressions to normal engine startup behavior.

How to test manually

1. Run the database and engine containers locally (see deployment/local.docker-compose.yaml).
2. Start the engine with the database already running — verify it starts without waiting the full timeout.
3. Start the engine before the database — verify it waits and starts successfully once the database comes up.
4. Start the engine with the database down — verify it exits with a clear error after the timeout.