Refactor data store reader + writer classes to extract utilities

[yousefmoazzam]

With the gradual introduction of padding, the data store reader and writer classes are picking up some extra functionality (notably, in #401).

Some of the padding-related functionality could be reused in other data sources/sink implementations (such as in StandardTomoLoader), so it would be useful to refactor the reader and writer classes and extract anything reusable.

[yousefmoazzam]

Some small refactoring was done in #414 (extraction of boundary extrapolation logic to a separate padding.py module), but it's possible that more refactoring can be done.