Chrome Version: (M71)
OS: (Win10)

The current method is we combined all damage rects in surface aggregater and then we separate (carve out) the video quads from the combined damage rects in ProcessForDCLayers. The better way is not to combine them at the first place. So we don't have to separate them again in dc_layer_overlay.cc.

This will improve the power of the underlay case since we don't have to redraw everything when the video is the only thing changed.