Create utility library to avoid redundancy in `pt2ts.py` across examples

[ma595]

There seems to be redundancy in the examples' pt2ts.py file. Could we consider refactoring the routines into a library and therefore condense each conversion in order to bring clarity to what's actually being done across the examples?

As well as conversion, pt2ts also checks the format of the model (by passing an example tensor) and prepares it for running on the desired architecture.

If I'm misunderstanding any underlying design choices, please feel free to close.

[jatkinson1000]

I really like this suggestion.
Could be a good hackathon group project?