Attempt to report unhandled errors as failures to cloud

[klutchell]

LGTM

[alexgg]

Manual tests against BM:

Problem reproduction:

• Trigger update and disconnect while image is being downloaded
• wait a couple of minutes
• Reconnect - allow engine to resume download
• Engine finishes download but reports error
• The update script hangs without exiting and reporting failure or removing locks

After the patch, the script correctly ends and reports failure.

Also tested:

• [x] If BM is not reachable to report failure, the script waits and keeps trying using curl's default backoff algorithm
• [x] Manually launching the script and interrupting it at different points correctly reports failure to cloud