Improve diagnostics: add logging and better error handling

[RKrahl]

Should improve diagnostics:

Add better error handling. Write a meaningful error message to stderr rather then letting Python emit the stack trace from the unhandled exception.

Add logging. Log relevant events such as zypper calls and return codes with level DEBUG. Log fatal errors (e.g. unhandled exceptions, see above) with level CRITICAL. Log to stderr with level DEBUG if it is a tty and to syslog with level WARN.