Add a delay flag for upgrading container clusters to support smaller clusters.
When upgrading a smaller container cluster (3 ~ 9) nodes there should be a --delay=timeinseconds option for the 'gcloud container clusters upgrade' command. In smaller clusters destroying and recreating nodes can kill some services in my cluster if the new nodes are not starting the pods as fast as new machines are destroyed and creating. Mostly due to docker pull's which take some time to complete, for example during the last upgrade it nearly got my database cluster down. A simple flag with a delay between node deletion and creation would solve this issue since I can specify a time that is safe for my cluster size.