Collision detection fails occasionally

[pedromartins]

Collision detection occasionally fails.

Steps to reproduce:

1. Go to a spot where you are surrounded by obstacles in one direction
2. Go back and forth in that direction
3. The ball eventually goes through one of the cubes

(Reported by Hok Poon)

[pedromartins]

My collision detection was initially calculated for uniform time steps (which would correspond to uniform distances travelled in the grid). When I made the animation time-dependent, the points that the ball passes through are more unpredictable, and that broke my assumptions for collision detection so I'll need to redesign it.

I'll do that when I get some free time. I'm surprised it actually works this well because it makes little sense with variable time steps.

[pedromartins]

"Fixed" by having a "snapping" integral, and a much more lenient collision detection that relies on the snapping. It should work in all cases, but it isn't elegant so I will try to redesign the whole system later.