Track Python-bindings improvements

[commial]

With #190, an initial Python support is provided.

It is still in its early days, and the API might change a lot.

This issue is used to track the futures improvements and associated limitations.

TODO

• for writing mode, .finalize can be called on Drop to allow for an easier usage.
  • ⚠️ .finalize can't be called twice, otherwise a MLA error will be raised. The Python API can wrap it on Drop, allowing the user to either call it or not
• support writing chunks to multiple file. A basic implementation would be the direct wrapping of start_file, add_file_content and end_file
• support for linear_extract wrapping. This could be made using a Python callback, or a given interface (as a Python abstract class) to derive
• add basic examples to a examples/ directory, and tests them
• add a README and document usage
• add build & test instructions (based on maturin and pytest)
• add a CHANGELOG to track changes
• add a version badge in the main README to track published version

Known caveats

• For now, MLA structures are not Sync. As a result, a reference on an archive cannot transit between Threads (but they are Send, so the struct itself can)
• A design change is likely required before having a Python object standing for a file in the Archive. Such an object would ease a more Pythonic interface, with a io.BufferedIO derived class.