Allow scale-to-zero for multiple instances in a service group

Currently, scale-to-zero is implemented as part of autoscale when the service group is scaled down to 0 instances. This means you cannot have multiple instances scaled-to-zero for the same service.

However, to reduce scaling latency to a minimum it would be preferable to have scale-to-zero as a property on the instance, completely independent of the periodic policy-based autoscale. This would allow you to have a set of replicas all in standby and immediately available when needed.