We need a way to farm deploy-hasingleton.

The current farm service isn't suitable to the task because it doesn't differentiate
the farm operation from the deploy operation.

To support this, we need an extension to the farm service that operates in two modes.
1) Farming only. Look for new deployments and distribute them across the cluster
2) Farming and Deploy. Deploy the farmed deployments.

This will allow the user to hot deploy to deploy-hasingleton from any node in the cluster.
But it will only actually be deployed on the master node (as required by deploy-hasingleton).