Add compile and run argument classes

[kudlicka]

• Introduce classes for compile and run arguments.
• Make compile arguments mirror tpplc arguments.
• Support saving and loading run arguments.
• Enable saving inference results to a file.
• Remove obsolete syntax highlighting in Jupyter notebooks.
• Refactor code for better PEP 8 compliance.

[kudlicka]

Examples

with treeppl.Model(filename="coin.tppl", m="smc-bpf", particles=100000) as coin:
    res = coin(outcomes=[True, True, False, True, False])

The treeppl.Model constructor arguments (except for filename and source) mirror the TPPL compile arguments, with dashes replaced by underscores.

The treeppl.Model now also supports a run method, so the previous example is equivalent to:

with treeppl.Model(filename="coin.tppl", m="smc-bpf", particles=100000) as coin:
    res = coin.run(outcomes=[True, True, False, True, False])

Saving the arguments of the last run of the model:

with treeppl.Model(filename="coin.tppl", m="smc-bpf", particles=100000) as coin:
    res = coin(outcomes=[True, True, False, True, False])
    coin.run_arguments.save_json("input.json")

Note: One potential improvement to discuss is moving the run_arguments property to an InferenceResult object, or possibly supporting both locations for better organization.

Loading arguments from a JSON file:

with treeppl.Model(filename="coin.tppl", m="smc-bpf", particles=100000) as coin:
    res = coin(arguments=treeppl.RunArguments.from_json("input.json"))

[elegios]

This passes tests and can be merged. Since it makes user-facing changes, could you summarize them briefly and put that in a PR to the changelog under the "unreleased" heading on the webpage (https://github.com/treeppl/treeppl.github.io/blob/master/docs/changelog.md) and then ping me?