Sending shutdown is handled gracefully in any case - as long as the UPS runtime is long enough. Backups will report "Task errors" after the fact, but that's to be expected, right?
Backup jobs won't restart "in the middle" after rebooting. There is no reboot-persistent state of such a canceled job.
They are started again per your schedule.
There will be some new data already transferred from PVE to PBS before the reboot, of course. Those chunks are not referenced in any intact backup and will be removed by the next garbage collection.
If you run a new backup before GC happens the chunks are then referenced by the new backup and will live. This way the already transferred data is not transferred again. (But probably it has to be read again on the PVE side as the dirty bitmap might be dropped - I am not sure about this one...)