* fix(pulse): keep the container recoverable when an update fails
- Do not delete /opt/pulse/pulse before the download. CLEAN_INSTALL=1
already wipes the target, so the rm only widened the failure window and
destroyed the v5 binary when the fetch failed.
- Guard fetch_and_deploy_gh_release and restart the running installation
when it fails, instead of exiting through the error handler with Pulse
stopped.
- Re-enable the unit after the pulse-backend to pulse rename, but only if
it was enabled before, so a deliberately disabled service stays that way.
* fix(pulse): roll back failed updates transactionally
* revert: keep Pulse update fix focused
---------
Co-authored-by: rcourtman <pulsemonitoringltd@gmail.com>
