Test, document and potentially improve reboot behavior after config changes

[eloquence]

We have at least one user report of being unable to log in after changing config.json with new onion service configuration and running sdw-admin --apply; a reboot resolved.

We should:

• Verify whether or not all VMs that need to be restarted are currently bounced as part of this process;
• If not, at minimum, document that a full system reboot is required;
• Potentially bounce said VMs as part of the --apply run.

[rocodes]

backlog pruning sesh:

• are we rebooting vms after config changes happen?
• part of our broader provisioning and/or install/update story
  see also/ related: [4.2] Write vm-specific config values to qubesdb #936