Improve backup error propagation and messaging

[jenoudet]

Why are the changes needed?

The backup command needs to be more responsive. These changes make a follower tasked with taking a backup that cannot currently do so (because it is restoring from a checkpoint) return an error to the leader immediately, rather than waiting for a timeout. This makes the bin/alluxio fsadmin backup command much more responsive and interactive.

Does this PR introduce any user facing changes?

No.