Continually try updating scaleset configs until successful

[bmc-msft]

As is, if a scaleset is "failed" for some reason (such as Azure VM allocation failed), the service does not attempt to push updated VM configs, which can cause our time-limited SAS tokens in the VM config to timeout.

This PR makes 2 changes:

1. attempts to perform any update as long as the scaleset isn't in the "Updating" state
2. moves the config update attempts from once per day configured once per day, but runs every 90 seconds until it's successful.