Force spuNodeCallback to be 16-byte aligned, so that we can use SIMD. Related to Issue 657

Recompute the quantization values in case the maximum/minimum extends roundtrip using quantize/unQuantize are out-of-range
(alternatively, pass in a larger quantizationMargin)

src/BulletMultiThreaded/SpuNarrowPhaseCollisionTask/SpuGatheringCollisionTask.cpp

@@ -291,7 +291,7 @@ SIMD_FORCE_INLINE void small_cache_read_triple(	void* ls0, ppu_address_t ea0,
 
 
 
-class spuNodeCallback : public btNodeOverlapCallback
+ATTRIBUTE_ALIGNED16(class) spuNodeCallback : public btNodeOverlapCallback
 {
 	SpuCollisionPairInput* m_wuInput;
 	SpuContactResult&		m_spuContacts;