Regions of interest

Think about how one could facilitate regions of interest. @ramav87 discovered that the HDF5 region references only support simplistic slices as opposed to arbitrary blobs. Perhaps regions could only be defined using USID given that regions would not necessarily be regular grids
Alternatively, we could collapse dimensions and do some book-keeping to facilitate tracking of regions of interest. The idea was that a region of interest would itself be a sidpy.Dataset object.

Here is the last message from the original issue in pyNSID"

Existing functions for region references in h5py is insufficient as it only allows for slice objects and not non-standard indexing (like lists of points). Therefore, an alternate route has been suggested which involves adding a region reference attribute to the NSIDataset, that could be setup as follows:

nsid_dataset.create_region() #dict('0':[0,5,3], '1':[2,10,15], 'capacitors') #pass a dictionary with the slice values and the name of the region, this should get stored
nsid.capacitors.points #gives back the dictionary used to make the region reference
nsid.capacitors.data #gives back the data

This is a TODO item, but it is low priority.

Start by reviewing h5py region references and the existing NSIDataset and sidpy.Dataset abstractions. Clarify whether regions should be represented through USID or dimension-collapsing bookkeeping, then define how region points and data are exposed. Done should support non-standard regions such as lists of points and provide a settled region API.