Better way to detect whether there are occlusion in plots for ranking

I _believe_ right now we only check if the plot is aggregated.

But if the plot is aggregated but have other detail channels (e.g., `detail`, `color`), they there might be occlusion.

In addition, for x:discrete/y:discrete plots if each pair of (x,y) has only one unique value / row, then there is no occlusion too. 