kubernetes: GCE kube-up doesn't install kube-proxy on the master

Using the standard cluster/kube-up.sh installation script on GCE, it looks like the kube-proxy is installed on every node but not on the master.

This is a problem because even though the master is marked as unschedulable, the DaemonSet controller will still run pods on the master node, and those Pods won’t be able to communicate via Kubernetes Services.

And example would be a DaemonSet which tries to talk to the k8s API, but fails on the master since the kubernetes.default Service ClusterIP is not programmed. Similarly, Pods started on the master via a DaemonSet won’t have DNS access.