Implement parallelized version of table calculation for large 2d segmentations

[constantinpape]

to support large (2d) image data. For this we need to extend cluster_tools.morphology (and the underlying nifty.distributed functionality) to also support 2d data.