plots: support two independent variables and a categorical dependent variable

[GraemeWatt]

For tables with two independent variables, where the values are either numerical or non-numerical, and all values of the dependent variable are non-numerical, it would be useful to have a suitable visualisation. This could be similar to a heatmap plot with a separate colour for each distinct value of the dependent variable. For example, it could be used to draw observed and expected exclusion contours at different confidence-levels on the same plot. Or to specify the best signal region in the mass plane of two hypothetical particles (see example).