json infer_schema drops null-only object keys

[cmdlineluser]

Checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of Polars.

Reproducible example

import polars as pl

data = b"""[
{"a": null, "b": 1, "c": null},
{"a": null, "b": 2, "c": null},
{"a": null, "b": 3, "c": null}
]"""

pl.read_json(data)
# shape: (3, 1)
# ┌─────┐
# │ b   │
# │ --- │
# │ i64 │
# ╞═════╡
# │ 1   │
# │ 2   │
# │ 3   │
# └─────┘

Log output

No response

Issue description

It seems that at least 1 non-null value is required for an object key to be included in the final output when using pl.read_json / .str.json_extract

From what I can tell, filter_map_nulls skips all nulls

polars/crates/polars-json/src/json/infer_schema.rs

Lines 78 to 83 in 89cc1e2

	fn infer_object(inner: &Object) -> PolarsResult<DataType> {
	let fields = inner
	.iter()
	.filter_map(|(key, value)| {
	infer(value)
	.map(|dt| filter_map_nulls(dt).map(|dt| (key, dt)))

Which results in the "all-null" keys being ignored/dropped?

Expected behavior

Just using pandas as a comparison, a and c are present in the result:

import io
import pandas as pd
import polars as pl

data = """[
{"a": null, "b": 1, "c": null},
{"a": null, "b": 2, "c": null},
{"a": null, "b": 3, "c": null}
]"""

pl.from_pandas(pd.read_json(io.StringIO(data)))
# shape: (3, 3)
# ┌──────┬─────┬──────┐
# │ a    ┆ b   ┆ c    │
# │ ---  ┆ --- ┆ ---  │
# │ f64  ┆ i64 ┆ f64  │
# ╞══════╪═════╪══════╡
# │ null ┆ 1   ┆ null │
# │ null ┆ 2   ┆ null │
# │ null ┆ 3   ┆ null │
# └──────┴─────┴──────┘

Installed versions

--------Version info---------
Polars:              0.19.9
Index type:          UInt32
Platform:            macOS-12.6.7-arm64-arm-64bit
Python:              3.11.6 (main, Oct  2 2023, 20:46:17) [Clang 14.0.0 (clang-1400.0.29.202)]

----Optional dependencies----
adbc_driver_sqlite:  <not installed>
cloudpickle:         <not installed>
connectorx:          <not installed>
deltalake:           <not installed>
fsspec:              2023.6.0
gevent:              <not installed>
matplotlib:          <not installed>
numpy:               1.25.0
openpyxl:            <not installed>
pandas:              2.0.3
pyarrow:             12.0.1
pydantic:            <not installed>
pyiceberg:           <not installed>
pyxlsb:              <not installed>
sqlalchemy:          <not installed>
xlsx2csv:            <not installed>
xlsxwriter:          <not installed>