Fix: libpe_status: calculate secure digests for unfencing ops

The calculation of digests for detection of when unfencing is needed reused
rsc_action_digest(). However that would only add secure digests when the
pe_flag_sanitized flag was set, which is only set by crm_simulate, so secure
digests would never be added in normal cluster operation. This led to
node attributes like name="#digests-secure"
value="stonith-fence_compute-fence-nova:fence_compute:(null),".

Now, rsc_action_digest() takes a new argument to select whether secure digests
are added, which is always set to TRUE when calculating unfencing digests.