[ODE] Colliider space/geom issues

[John Miles]

At line 497 of collision_quadtreespace.cpp in the current unstable snapshot,
someone with the initials (mg) has added a comment "there can be multiple
instances of a dirty object on stack be sure to remove ALL and not just
first, for this we decrement i."

A newly-added line below that comments reads "--i;", just before a break
statement exits the loop.  However, i is never used anywhere outside the
loop, so the decrement operation has no effect.

Was this intended to fix a bug, or...?

Also: is anyone planning to update the dTerrainX/dTerrainY and dCone
collider contributions to support the new geom-offset functionality?  It's
well past time to move those geometry types to the main branch, IMO.
They're reasonably well-supported by the other geometry types and very
reliable in my experience.

-- jm