Autoscaling & scale to zero

Every 3s Aloof looks at in-flight requests per instance (and optionally CPU %), and scales between min and max. Scale-down waits 15s.

aloof autoscale api --min 1 --max 6 --concurrency 20 --cpu 70
aloof autoscale preview --min 0 --max 2 --idle 120    # scale to zero after 2 idle minutes

With min 0, the first request after idle is held at the edge while an instance starts (about 1s for a small CPU app).