Fix: fencing: log all peers' results consistently

This refactor functionizes adding one peer's results to an operation's result
list, mainly for readability (especially after more code is added for remapping
reboot to off+on).

It does make one exceedingly trivial logging improvement. Previously, peers
that reported no devices were handled separately. Now, the addition of the
result is conditional on there being devices, so otherwise all results go
through the same logic and get the same log messages.