[ODE] rolling friction

[Nagymathe Denes]

>> Using angular damping may look correct in most situations, but physically 
>> it
>> isn't (at least you have to check whether the body is touching some 
>> surface
>> or not to avoid the damping in mid-air).
>
> Not sure if this is still an issue but objects with certain inertia 
> tensors used to gain momentum when spinning in the air. Thus applying some 
> angular damping even if nothing is being touched would not be a bad idea.

It's true, i agree, but it's caused by some numerical inaccuracy in the 
angular integration. Thus it should be handled there (in dxStepBody()), and 
cannot be too strong. Some do some sort of 'clamping' the angular velocity 
to some reasonable upper limit, and leave the bodies rotating free below 
that. I'd suggest using some small 'default' angular damping in dxStepBody, 
and a separate rolling friction.

Denes