Ray collision testing spuriously failing

Raycasting against 2D shapes such as `ConvexPolygon` and `Polyline` are failing spuriously with very small shifts in the origin of the ray. The problem appears to be something with the raycasting algorithm itself. I tested with both `f32` and `f64` for the data structures used in case it was something to do with floating point accuracy, but the problem persisted for both.

I implemented a failing test in #219 that demonstrates the problem in a simple scenario.