There are discontinuities in the velocity field at cube face boundaries

[smcameron]

There are discontinuities in the velocity field at cube face boundaries. These lead to sharp corners in the fluid motion at face boundaries. This is likely due to abruptly switching which axes are used to calculate the noise gradient at face boundaries. However, changing the code to always use the same axes did not seem to fix it (presuming I did it right -- in which case it causes other problems -- also possible I didn't do it right -- too many rotations of the coord system and my brain goes all melty.)