Feature: fencing: improve exit reason for fencing timeouts

Troubleshooting timeouts is one of the more difficult aspects of cluster
maintenance. We want to give as much of a hint as possible, but for fencing in
particular it is difficult because an operation might involve multiple retries
of multiple devices.

Barring another major project to track exactly which devices, retries, etc.,
were used in a given operation, these changes in wording are probably the best
we can do.