Geom.density2d with color grouping

[Mattriks]

This works on Gadfly master:

using RDatasets
using Gadfly.CategoricalArrays: cut
quakes = dataset("datasets", "quakes")
quakes.depthg = cut(quakes.Depth, 3)

lyrs = [layer(df, Geom.density2d(levels=4), color=df.depthg[1:1])  for df in groupby(quakes, :depthg)]
theme1 = Theme(point_size=1.2pt, alphas=[0.5], discrete_highlight_color=c->nothing)

p1 = plot(quakes, x=:Long, y=:Lat, color=:depthg, reverse(lyrs)..., Geom.point,
    Scale.color_discrete_manual("deepskyblue","hotpink", "yellow3"), theme1)

But it would be nice if Geom.density2d auto-grouped by color:

p2 = plot(quakes, x=:Long, y=:Lat, color=:depthg, Geom.point,
    layer(Geom.density2d(levels=4), color=:depthg), theme1)

[tlnagy]

Given that we auto-group by color for most everything else, this would be a great quality of life improvement. I recently ran into this problem when trying to simplify extremely dense scatter plots.