The TEMPO and PT-TEMPO backend is hard to read

[gefux]

The actual tensor network code underpinning the TEMPO and PT-TEMPO implementations is way to hard to follow. It would be good to tidy up the code and leave a few (but not too many) helpful commands.

[gefux]

Also, since we only have a single backend (using google's TensorNetwork), we should flatten out (remove) the backend structure and lift the computation into tempo.py and pt_tempo.py