Rethink how testing is done in zppy & zppy-interfaces. zppy should simply test that it can invoke zppy-interfaces global-time-series. zppy-interfaces however should test that Global Time Series plots actually generate correctly.
Maybe we could have zppy test that zppy is passing in the right parameters and then have zppy-interfaces actually do the image comparisons of the plots, using pre-generated ts and mpas_analysis output.