Deployment fails if OpenSUSE repos are unavailable

[juhaj]

Deployment will fail if any of the zypper commands fail; likewise if any of the pip commands fail. This is particularly problematic as it may affect a single node only: if it was a problem with either all or none recreating the cluster would be an ok workaround but I don't like recreating the whole thing just because two nodes out of 40 fail.

Options:

1. Do not run the azuredeploy.sh as part of the deployment but rather by hand from master after everything else has succeeded (cron? with a check of whether the node has already run it to avoid reruns).
2. Gracefully retry failing commands in azuredeploy.sh: this is not very nice since we'd either need to be able to retry everything or find out all the potentially failing commands.