Fix test failure caused by making genconfig trackerless When I removed the need to have a tracker for genconfig, I broke a couple of tests. The breakage was hidden by a bad workflow. This change allows the tests to pass and refactors reporting of UsageError exceptions in admin.py. It also adds a test for failure in install command as there is a test case for it and this test tests the refactor.

#! /bin/sh

# if there are multiple trackers, d=demo t=tracker ...
# returns last one for testing that server is up. Does not test
# each tracker.
tracker=$(ps -ef | sed -ne '/roundup-server/s/^.*\s\(\w*\)=.*$/\1/p')
wget -q -O /dev/null --no-verbose http://localhost:8080/${tracker:-demo}/