these have moved into functions of ModelWrapper, they weren't
really proper transformations anyway

currently wrapper for ONNX shape inference, will change later

this will be eventually performed in ModelWrapper automatically
when needed

the intention is to use instances of this class to represent
ONNX model instances, and gradually add useful utility functions.

complete with all intermediate tensor values. it renders a bit
strangely in netron atm but still useful for debugging.