Putting the command line argument handling into the central test
context seems like the better solution, in particular considering that
argument handling might get changed in the future to use Viper.