[DRAFT] Move adaptive scaling functionality into dedicated process

**Problem**
With adaptive scaling being executed on the client-side, it's easy to execute multiple (possibly mutually-exclusive) adaptive scaling policies on a single distributed cluster through the use of multiple `Cluster` objects.

**Goal**
At any given moment, there should be one, and only one, adaptive scaling policy in place that is being applied to the distributed cluster (not `Cluster` object).

Additional benefits:
* simplifies the client-side
* allows clients to query and modify the adaptive scaling policy another one has put into place