feat(python): Parse JSON data in Utf8 to polars dtype

[josh]

I was looking for a native replacement for a simple apply(json.loads) UDF that also worked well on lazy frames. I saw str.json_path_match but I really wanted a parsed struct (or whatever dtype) back, not a string value.

It looks like some initial work on this started back in #3413 and got partially exposed in #5140. A private Utf8Chunked.json_extract helper was added, but it never was fully exposed publicly on the Py Series or Expr APIs. So this PR exposes it.

The API optionally supports dtype inference on eager frames and series. When used on a lazy frame, the default unknown dtype will properly lead to an error. Additionally, a nice feature over apply(json.loads) is that a partial dtype can be supplied to omit struct keys you're not interested in decoding. Seems to be a nice property of using read_ndjson under the hood.

How does the name json_extract still sound? That's the existing name of the internal function so just went with it. parse_json might have been my first instinct, but deferring to the maintainers preference.

[ritchie46]

Thanks a lot! That's great functionality indeed. Can you expose the methods to the docs (on alphabetical order).

[josh]

Can you expose the methods to the docs (on alphabetical order).

✅ Done