Let the testing interactor take care of the image comparisons when
appropriate. Currently has a hard coded tolerance of 40, which should
work for most of the chart tests.