Use separate collision layers for input/button actions vs physics

[sorenhan]

input (hover+click) runs 100% locally, and is a perfect candidate to be controlled with group masks (and maybe also be tied to appearance enabled flag) Rigid body physics on the other hand is synchronized, and would become very unpredictable if it was group maskable.

Considering that you often want to treat these cases independently, we should consider splitting collision for these two cases. For unity, that would mean requiring an additional layer to manage the collision.