feat(rust, python): Implement schema, schema_override for pl.read_json with array-like input

[andysham]

This PR fixes "ComputeError: "schema arguments not yet supported for Array json", see #11505

The following operations are now supported, (taken from unit tests)

def test_to_from_buffer_arraywise_schema() -> None:
    buf = io.StringIO(
        """
    [
        {"a": 5, "b": "foo", "c": null},
        {"a": 11.4, "b": null, "c": true, "d": 8},
        {"a": -25.8, "b": "bar", "c": false}
    ]"""
    )

    read_df = pl.read_json(buf, schema={"b": pl.Utf8, "e": pl.Int16})

    assert_frame_equal(
        read_df,
        pl.DataFrame(
            {
                "b": pl.Series(["foo", None, "bar"], dtype=pl.Utf8),
                "e": pl.Series([None, None, None], dtype=pl.Int16),
            }
        ),
    )


def test_to_from_buffer_arraywise_schema_override() -> None:
    buf = io.StringIO(
        """
    [
        {"a": 5, "b": "foo", "c": null},
        {"a": 11.4, "b": null, "c": true, "d": 8},
        {"a": -25.8, "b": "bar", "c": false}
    ]"""
    )

    read_df = pl.read_json(buf, schema_overrides={"c": pl.Int64, "d": pl.Float64})

    assert_frame_equal(
        read_df,
        pl.DataFrame(
            {
                "a": pl.Series([5, 11.4, -25.8], dtype=pl.Float64),
                "b": pl.Series(["foo", None, "bar"], dtype=pl.Utf8),
                "c": pl.Series([None, 1, 0], dtype=pl.Int64),
                "d": pl.Series([None, 8, None], dtype=pl.Float64),
            }
        ),
        check_column_order=False,
    )

[korommatyi]

This partially solves #11148 as well.

[ritchie46]

Nice, thanks!