tox: set ignore_errors
    
    This doesn't actually *ignore* errors, it just means we continue
    running commands after one errors out instead of immediately
    bailing. This avoids problems like "the tests don't run if
    check-needles fails, so then the coverage reports fail because
    there's no coverage.xml".
    
    Signed-off-by: Adam Williamson <awilliam@redhat.com>