False values from CategoricalSeries.unique()

[s-banach]

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

df = pl.DataFrame({"x": [str(n) for n in range(50)]}).cast(pl.Categorical)
print(df.unique().null_count())

shape: (1, 1)
┌─────┐
│ x   │
│ --- │
│ u32 │
╞═════╡
│ 2   │
└─────┘

Log output

No response

Issue description

Notice that nulls appear in a series without nulls, after taking unique().
Sometimes python quits without printing anything, depending on the categorical series I test with.
Bug does not seem to be present in version 1.10

Expected behavior

Don't add nulls, don't crash.

Installed versions

--------Version info---------
Polars:              1.11.0
Index type:          UInt32
Platform:            Windows-11-10.0.22631-SP0
Python:              3.12.2 | packaged by conda-forge | (main, Feb 16 2024, 20:42:31) [MSC v.1937 64 bit (AMD64)]
LTS CPU:             False

----Optional dependencies----
adbc_driver_manager  <not installed>
altair               <not installed>
cloudpickle          <not installed>
connectorx           <not installed>
deltalake            <not installed>
fastexcel            0.10.4
fsspec               2024.6.1
gevent               <not installed>
great_tables         0.9.0
matplotlib           3.8.4
nest_asyncio         1.6.0
numpy                2.1.0
openpyxl             3.1.5
pandas               2.2.2
pyarrow              17.0.0
pydantic             2.9.2
pyiceberg            <not installed>
sqlalchemy           2.0.35
torch                <not installed>
xlsx2csv             <not installed>
xlsxwriter           3.2.0

[cmdlineluser]

I haven't been able to get a crash, but do get a null_count() of 2 each time.

Just some debugging notes:

It seems to have changed after #19359

It also seems to go away if I set POLARS_MAX_THREADS=1

[ritchie46]

@orlp can you take a look?

[orlp]

I can reproduce, but only with POLARS_MAX_THREADS=2.