MPI inversions with blocks of pixels

[aasensio]

Currently, MPI inversions/synthesis are done in a pixel-by-pixel mode. For inversions this is OK because every pixel takes a long time. However, doing that for synthesis is inefficient because the synthesis takes much less time. An option to send blocks of pixels should be available