Bugfix/EdgeCarryShift

[JonathanMaes]

While the newly added EdgeCarryShift (proposed in #316) worked as mathematically expected, the original implementation allowed nonphysical situations to arise where the zero magnetization (0,0,0) from edge cells outside the geometry was copied to cells which, after shifting, lay inside the geometry.

To address this, cells which would have been assigned the zero-vector, but are actually inside the geometry, are instead assigned the vector ShiftMagL, ShiftMagR, ShiftMagD or ShiftMagU (depending on the shift direction).

Observe an example of the updated EdgeCarryShift behavior below.
A slanted geometry with random initial magnetization is shifted 32 cells to the right. As EdgeCarryShift = true, the left edge values are applied to all the newly introduced cells. For the cells at the very bottom of the new geometry, those left edge cells were outside the geometry, so the vector ShiftMagL is applied (set to (-1,0,0) here). The pre-bugfix behavior was that these bottom cells would have magnetization (0,0,0) (grey).

I also implemented the geom.GetCell() method, which allows a sanity check in the updated test/edgeCarryShift.mx3 test script.