Improve failure handling

[wgwoods]

Basically, when the system crashes, we should probably display an error message and tell the user how boned they are, and let them choose between a shell and rebooting.

More specifically - if the upgrade-prep service fails, we should:

• print a message about the failure
• log failure to /var/log/upgrade.log
• do the "root password for debugging or reboot" prompt
• reboot if no activity after ~60 sec

[wgwoods]

See https://bugzilla.redhat.com/show_bug.cgi?id=910086

[wgwoods]

We might also want fedup to provide info about failures via fedup status or fedup log - see #51 for more thoughts on commandline/workflow stuff.