Fix maintenance duration with delays

[Jorres]

Account for both --delay-between-restarts and --nodes-inflight when sending the request to CMS. Worst case scenario, CMS will give you all nodes that you asked for, which means that in total the duration must be:

1. Determine base restart duration: reasonable single restart duration * reasonable retry number (currently, 60 and 3)
2. Determine how many batches it will take (total requested nodes / --nodes-inflight)
3. Total result will be: number of batches * (base restart duration + --delay-between-restarts)