Log: controller: improve action result messages

Previously, processing a graph event could produce a mishmash of several
obscure log messages. Now, there is one log message per event (at notice or
info level, because the scheduler will log an error or warning if appropriate),
and the messages for different results are consistently formatted.

This allows the code to be significantly simplified as well.

Example:

warning: Action 5 (FencingPass_monitor_0) on node2 failed (target: 7 vs. rc: 0): Error
info: Action FencingPass_monitor_0 (5) confirmed on node2 (rc=0)
info: Detected action (32.5) FencingPass_monitor_0.22=ok: failed

->

notice: Transition 32 action 5 (FencingPass_monitor_0 on node2): expected 'not running' but got 'ok'