Saner JSON formats

[akissinger]

Creating a PR before I merge, since this is potentially a breaking change.

Current JSON formats for proofs and rewrites have JSON containing strings containing JSON (which sometimes contain strings containing JSON containing strings containing JSON...).

Technically this works, but its a bit clunky to run the JSON parser many times to load a single file, especially if you want to consume JSON in an external tool. This branch gets rid of the nested JSON strings by splitting the old to_json method into two parts: to_dict which serialises to a python dict, then a thin wrapper to_json which just called json.dumps on the output of to_dict. Nested structures are added directly as dicts rather than strings.

The from_json will either accept a string (in which case it parses JSON first) or a python dict. Hence, ZX live can still read old-style files, but it will save in the new format.

Note this needs zxcalc/pyzx@aab09ad .