Fix mult quarternion bug

[mororo250]

There was a bug in the mulQuat multiplication We were doing:
vz + uv_z * w2 + (qz * uv_y - qy * uv_x) * 2 instead of vz + uv_z * w2 + (qx * uv_y - qy * uv_x) * 2

I also optimized the function to use SIMD vectorized methods.