[ODE] Triangle-box collider contribution

[Adam D. Moss]

Russ Smith wrote:
> my own opinion is that the core code should be robust, have a
> maintainer, and be well documented - not necessarily that all of the
> O(N^2) collision cases be supported.

I'll continue to be more reluctant, however, when the collision
cases aren't all implemented.  I just can't easily swallow the idea
that dropping a selection of core ODE primitives into a room and
having certain combinations simply fail to collide is an acceptable
state of affairs, at least now that we perhaps draw close to ODE 1.0
(which incidentally I think would recede faster than it approaches
if we were to be any looser about what we absorb into the core in
an incomplete state).

The cylinder code I'm not too worried about, since basically
we have an almost-fully-formed set of colliders offered by Croteam,
ray-vs-cylinder is already in the dCylinder contrib, and some
people are actively keen on doing the necessary porting (and
implicitly some degree of maintainance, I hope :)).

Regards,
--Adam
-- 
Adam D. Moss   . ,,^^   adam at gimp.org   http://www.foxbox.org/   co:3