In cleanup commit 9cd05a8c39c6ed03688859de92372d93af5cc6e2 the labels on the ServiceMonitors were changed from k8s-apps to k8s-app. This commit makes the Prometheus selector match that change.

Whenever serving Prometheus or Alertmanager through non root
ExternalURLs we require the reverse proxy to trim the URL to be to the
root Prometheus/Alertmanager.

* namespace has to be kube-system

In order for this service to be discovered by ServiceMonitor k8s-apps-http, it should belong to namespace kube-system.

* namespace has to be kube-system

* namespace has to be kube-system

for service discovery to work within prometheus configuration (ServiceMonitor k8s-app) these service and endpoints should belong to kube-system namespace. Otherwise the service monitor won't fetch it.

* selector was incorrect

selector prometheus: prometheus-frontend is incorrect, the name of the Prometheus resource in the example is "frontend", not "prometheus-frontend"

Rename kube_pod_container_requested_memory_bytes -> kube_pod_container_resource_requests_memory_bytes in grafana dashboard

Write an example for basic auth

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

Now that we have the possibility to use the honor_labels configuration
there is no need for this specially treated annotation anymore.