Update needs to take care of archiving existing log and dump files

[askfongjojo]

In the current update manual procedures (and programmatic update in the future), unarchived logs and dumps are gone when zones are removed. We should include a step to invoke the archive jobs beforehand.

[bnaecker]

I've read briefly through the proposed update procedures in GDrive, and this may be alleviated somewhat by the recent zone-bundling features. In particular, if we implement the update by asking the sled agent to stop all VMs, that will trigger collection of a zone bundle from the Propolis zones. That bundle includes all log files, current, rotated, and those archived to the U.2. This doesn't yet apply to the sled agent shutting now service zones, but could be made to do so pretty easily.

This isn't a substitute for more careful archiving of the logs before performing an update, but it may mean we can defer some of that work if needed. Also note this doesn't include any core files or crash dumps.