[cern] [k8s] Monitoring with Prometheus and Grafana

Cherry-pick: https://review.openstack.org/#/c/426291/

Profit from the default cAdvisor deployed by k8s to deploy the remaining monitoring stack on top, made of node-exporter, Prometheus and Grafana.

Node-exporter is ran as a normal pod through a manifest, while Prometheus and Grafana are deployments with 1 replica.

Prometheus has compliance with Kubernetes, so the discovery of the nodes and other k8s components is configured directly in Prometheus configuration.

Change-Id: If2cab996b9458580a55b5212ab298c909622e7f3 Partially-Implements: blueprint container-monitoring

Conflicts: magnum/drivers/common/k8s_template_def.py magnum/drivers/k8s_fedora_atomic_v1/templates/kubecluster.yaml magnum/drivers/k8s_fedora_atomic_v1/templates/kubemaster.yaml magnum/drivers/k8s_fedora_ironic_v1/templates/kubecluster.yaml magnum/drivers/k8s_fedora_ironic_v1/templates/kubeminion_software_configs.yaml magnum/tests/unit/conductor/handlers/test_k8s_cluster_conductor.py

Closes #4373