The idea of this grain boundary sliding approach is simply to set "boundary unodes" before running elle2fft. These unodes will be set to a 2nd phase which is the "boundary-phase" (isotropic and n-times softer than e.g. ice basal plane). 
Afterwards, we will run elle2fft, the fft code itself and then:

a) Run fft2elle and afterwards reset "boundary unodes" to real ice unodes
b) Reset "boundary unodes" to real ice unodes and run fft2elle

The user needs to decide on option a or b. Also, probably another routine is needed to reset boundary unodes to ice unodes than the one implemented right now.
