K8s 1.16 compatibility

[betermieux]

A first take on making CLUO ready for k8s > 1.16

• Changed 'ExtensionsV1beta1().DaemonSets' to 'AppsV1().DaemonSets'
• Updated packages 'k8s.io/apimachinery', 'k8s.io/client-go' and 'k8s.io/kubernetes' to stable 1.16.2 versions
• Needed to update some more dependend packages (also to current stable versions)
• There were major changes to the watch interface (moved from apimachinery to client-go). I tried my best to adapt the source files, but I am unsure about: https://github.com/betermieux/container-linux-update-operator/blob/2f5fe31992300915154c81ffcadbb731eae4d45c/pkg/operator/operator.go#L266 I have tested theses changes on a 1.16.2 cluster and everything seems to be working like before. Some more tests regarding leaderelection seem to be necessary.

[betermieux]

#196

[dghubble]

Yep, with Kubernetes v1.16+ the CLUO agent drain loops, since the API change means it doesn't exclude itself as it did previously. This update makes sense, its in line with the last time drain logic had to update for Kubernetes v1.9. Looks good to me.

However, myself and the other folks who used to approve and release aren't with CoreOS now. Perhaps if this was preceded with a bump to the Go version to fix the Travis setup, a Red Hatter could approve.