Testing on intensive variables might need more granularity

At present, we are testing against a [sum]( https://github.com/pysal/tobler/blame/45b9673863d6fcd751ece372024045c38e346b05/tobler/tests/test_area_interpolators.py#L37) which may miss invalid estimates for particular target areas. 