Workflow: scale-to-zero cost optimization for dev environments

# Workflow: scale-to-zero cost optimization for dev

## Goal
Dev/staging databases that cost near zero when idle and wake without errors when used.

## Steps
1. For each dev/staging compute, confirm scale to zero is enabled with the default 5-minute idle timeout (or set it explicitly).
2. Set autoscaling min to the smallest size (0.25 CU) so even active dev usage is cheap.
3. Align client pools with the idle timeout: SQLAlchemy `pool_recycle` at or below the timeout with `pool_pre_ping=True`; Django `CONN_MAX_AGE=0` or `CONN_HEALTH_CHECKS`.
4. Use schema-only or anonymized branches for dev data so a dev database contains no production PII worth protecting at rest beyond the defaults.
5. Review the bill after one cycle: compute hours for dev branches should collapse.

## Ordering constraints
- Set pool recycling before relying on wake behavior, or the first query after idle fails on a dead pooled connection and the "optimization" looks like an outage.
- Anonymize before handing the branch to developers, not after.

## Traps
- Do not apply this to latency-sensitive production; the cold start is the price of the savings.
- Cron jobs that run every 10 minutes keep the compute warm all day; either accept the cost or batch the jobs.
- Each branch's compute bills separately; ten dev branches each with a warm compute is not "dev is cheap".

## Verify
Console shows dev computes Idle most of the day; first-query wake succeeds; the bill's compute line drops.