Bump prometheus to 2.0.0, prometheus-operator to 0.15.0, alertmanager to 0.12.0 and node-exporter to 0.15.1, grafana to 4.6.3
migrate prometheus alerts to yaml notation

NAMESPACE feature in scripts implies it works. warnning ensures
that users are aware that its use requires further changes.

Alleviates #765

Per issue 593, append .yml to each filename used in the generated configmap. I wasn't able to see alerts in the Prometheus UI until I did this

The merge was missing a new line at the end of the merge, so only
the first datasource was being put into the ConfigMap.

In the generated version of this file at /contrib/kube-prometheus/manifests/grafana/grafana-deployment.yaml the version was bumped to 4.5.2 but seems it was overlooked in this templated version of the file.

https://github.com/kubernetes/kubernetes/issues/29542 has been fixed,
use apply instead of create to create third party resources in custom service
monitoring example.

Starting with Kubernetes 1.6 `kubectl apply` works for TPRs so we can
start using it, therefore using multiple files is no issue.

Without this, a race condition may occur, preventing grafana-watcher
from logging in to grafana.

These manifests should live in kube-system; otherwise the label selector
won't work.

Datasource links were duplicated in the grafana dashboads. This now also
allows exporting grafana dashboards from the UI and just dropping them
into the assets directory and they will be wrapped by the manifest
generation script.

For `--dry-run` to work with kubectl a Kubernetes cluster's apiserver is
actually used, which is unnecessary for generating these manifests. This
approach also allows further customization, such as adding labels to the
generated manifests.