[ODE] 0.9-rc1

[Irrisor]

Hi, just want to give more background for the reason I'm posting about that 
assert in dNormalize:

Irrisor wrote:
> The condition in dNormalize4 _can_ be false! E.g. when called from dxStepBody in
> dxQuickStepper after SOR_LCP has computed NaN values for cforce.
I have some apps running into this issue, and as I could not track it down to 
contact generation or bug in the lcp solver I circumvent it by moving the body a 
tiny little bit if NaN is the result. With an dAssert instead of instead of 
dDebugMsg in dNormalize4 I cannot recover from that ODE internal error any more.

/Irrisor