In lieu of #283, implement Stackdriver Canary Analysis for use in GKE.

Please describe the problem you have

klogv2 has the https://godoc.org/github.com/kubernetes/klog#SetLogger API, which should let us drop the vaguely sketchy output stream we use to capture core Kubernetes log output.

Currently this system's modes are changed only when contours are appeared above the icon area but it will not always be best idea. So if anyone can train a Gesture Recognition model and then assign a mode then this will add real aid.
Thanks.

Currently, the dex and gangway config files must be updated manually.

Automate this replacement. Perhaps use a templating system, ideally something less heavy-weight than helm.

Please describe the problem you have
Each e2e test creates a sample workload app (i.e. kuard) to test e2e functionality of a contour environment deployed by the operator. Each test should remove the resources, i.e. Ingress, used to deploy the app.