`sdw-admin --configure` should handle failures more gracefully

[eloquence]

Right now, sdw-admin --configure fails with a traceback on potential failure modes:

• files can't be found in the expected location
• vault VM fails to start (less likely, but we should still cover it)

We should handle these and other plausible cases more gracefully, ideally with useful output to help resolve them.

In addition, --configure does its work inside an exception handler clause, which means that any printed errors will be extra verbose ("while handling the error that the config doesn't exist, we encountered another error"). IMO we should move the logic outside the handler once we've determined the config file has to be created.

(My first read of the error was that config.json had to be manually created, so this kind of complicated error chain could also lead to user error.)