[ODE] Large trimesh triangles

[Geoff Carlton]

I've never seen this.  I have seen the opposide though, when a flat area 
is subdivided into many polys.  In this case, there seems to be some 
strange artifacts in both the ccylinder and the box - its mainly the 
geoms catching on tri edges when it shouldn't (needs adjacency).  I've 
actually done a few hacks that improved my particular case by either 
ignoring the improper normal contacts (when they are opposite to the 
triangle normal) or lerping them towards the triangle normal.

Geoff