[ODE] Triangle-box collider contribution

[Jeffrey Smith]

> The biggest problem (really, the only problem) I've had with Erwin's current
> rewritten solution is fall-through of (mostly) small, fast-moving objects.

Unfortunately, that is not the only (or even the most serious) problem with
Erwin's code.  As I've mentioned here before, it fails horribly if the
trimesh objects being collided are smaller than the bounding boxes.
Both my and asko's code address this bug.

The problem you're talking about is, basically, insoluble in the general case
unless we use swept volumes, or some clever time-decimation scheme for finding
the exact time of contact.  However, the workaround for these problems is
simple: decrease the timestep or decrease the velocities.

-Jeff