select contiguous data splits for test and train

[squassina]

Currently the prepare_dutchf3.py selects the edge of the cube as test and the core as train.

It is important to train with all parts of the cube so when we run the tests/validation we have more diverse data

[maxkazmsft]

Rescoped: right now train set starts from 0 always, and we want to have the ability to start from any index to be able to select any contiguous part of the data as train set, test 1 and test 2 sets. For example, if original dimensions are 100x100x100 (x, y, z), we can have train go from [20-80]x[20-80]x[0-100] and the rest be split into test1 and test2 sets by crosslines and inlines