Move Monitoring Service off Head Node

The first version of the Monitoring service, which is Ceilometer-based, runs on the POD's head nodes. So that we can minimize the security thread of collecting meters from untrusted sources (e.g., VMs/containers running VNFs), yet provide a unified monitoring framework for both infrastructure and tenant services, we want to move Ceilometer to a scalable number of compute nodes. A high-level design is available *here*.