[ODE] trimesh and capsules bug fix - swapping g1 and g2 in returning contacts

[Paul MacKenzie]

Jaroslav Sinecky wrote:
> There was another old mystery with trimesh and capsules. I always had to 
> swap the contact normal and didn't know why if it obviously worked in 
> ode test and for other users too. Today finally I sorted that out: the 
> g1 and g2 fields in geom struct were swapped when returned from 
> trimesh-capsule collision. Depending on whether you used or not this 
> information for some further processing and contact joint creation, it 
> appeared as a bug or not.
> 
> It's fixed and checked-in svn trunk. So if some had to swap the normal 
> also (I remember someone commenting  that some time ago), remember to 
> swap back now.

Good catch, Jaroslav.  My capsule geometries were accelerating through 
the mesh on contact, now with this change, they no longer do so.

Thanks very much!

Cheers,
Paul