Splitting master/node services into separate charm layers

This branch includes a rollup series of commits from a fork of the
kubernetes repository pre 1.5 release because we didn't make the code freeze.
This additional effort has been fully tested and has results submit into
the gubernator to enhance confidence in this code quality vs. the single
layer, posing as both master/node.

To reference the gubernator results, please see:
https://k8s-gubernator.appspot.com/builds/canonical-kubernetes-tests/logs/kubernetes-gce-e2e-node/

Apologies in advance for the large commit, however we did not want to
submit without having successful upstream automated testing results.

This commit includes:

 - Support for CNI networking plugins
 - Support for durable storage provided by ceph
 - Building from upstream templates (read: kubedns - no more template
 drift!)
 - An e2e charm-layer to make running validation tests much simpler/repeatable
 - Changes to support the 1.5.x series of kubernetes

Additional note: We will be targeting -all- future work against upstream
so large pull requests of this magnitude will not occur again.

cluster/juju/layers/kubernetes-worker/debug-scripts/inotify

@@ -0,0 +1,8 @@
+#!/bin/sh
+set -ux
+
+# We had to bump inotify limits once in the past, hence why this oddly specific
+# script lives here in kubernetes-worker.
+
+sysctl fs.inotify > $DEBUG_SCRIPT_DIR/sysctl-limits
+ls -l /proc/*/fd/* | grep inotify > $DEBUG_SCRIPT_DIR/inotify-instances

cluster/juju/layers/kubernetes-worker/debug-scripts/kubectl

@@ -0,0 +1,13 @@
+#!/bin/sh
+set -ux
+
+alias kubectl="kubectl --kubeconfig=/srv/kubernetes/config"
+
+kubectl cluster-info > $DEBUG_SCRIPT_DIR/cluster-info
+kubectl cluster-info dump > $DEBUG_SCRIPT_DIR/cluster-info-dump
+for obj in pods svc ingress secrets pv pvc rc; do
+  kubectl describe $obj --all-namespaces > $DEBUG_SCRIPT_DIR/describe-$obj
+done
+for obj in nodes; do
+  kubectl describe $obj > $DEBUG_SCRIPT_DIR/describe-$obj
+done

cluster/juju/layers/kubernetes-worker/debug-scripts/kubernetes-worker-services

@@ -0,0 +1,13 @@
+#!/bin/sh
+set -ux
+
+for service in kubelet kube-proxy; do
+  systemctl status $service > $DEBUG_SCRIPT_DIR/$service-systemctl-status
+  journalctl -u $service > $DEBUG_SCRIPT_DIR/$service-journal
+done
+
+mkdir -p $DEBUG_SCRIPT_DIR/etc-default
+cp -v /etc/default/kube* $DEBUG_SCRIPT_DIR/etc-default
+
+mkdir -p $DEBUG_SCRIPT_DIR/lib-systemd-system
+cp -v /lib/systemd/system/kube* $DEBUG_SCRIPT_DIR/lib-systemd-system