.backward_fill() does not consider np.nan to be invalid

[marc-at-brightnight]

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 numpy as np

df = pl.DataFrame({
    "f": [1.0, None, 1.0],
})

bfilled_df = df.select(pl.all().backward_fill())

assert bfilled_df[1, 'f'] == 1 # works

df = pl.DataFrame({
    "f": [1.0, np.nan, 1.0],
})

bfilled_df = df.select(pl.all().backward_fill())

assert bfilled_df[1, 'f'] == 1 # does not work

Log output

Traceback (most recent call last):
  File "/Users/mnhmbp/PycharmProjects/poweralpha/venv/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3577, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-4-cc1558aaef03>", line 17, in <module>
    assert bfilled_df[1, 'f'] == 1 # does not work
AssertionError

Issue description

.backward_fill() works as expected with Nones. However, with np.nan values, it does not work. This seems strange to me.

Expected behavior

It should work with np.nan as it does with None

Installed versions

--------Version info---------
Polars:               0.20.18
Index type:           UInt32
Platform:             macOS-14.2-arm64-arm-64bit
Python:               3.10.13 (main, Aug 24 2023, 12:59:26) [Clang 15.0.0 (clang-1500.0.40.1)]
----Optional dependencies----
adbc_driver_manager:  <not installed>
cloudpickle:          3.0.0
connectorx:           <not installed>
deltalake:            <not installed>
fastexcel:            <not installed>
fsspec:               2024.2.0
gevent:               <not installed>
hvplot:               <not installed>
matplotlib:           3.8.3
nest_asyncio:         <not installed>
numpy:                1.23.0
openpyxl:             3.1.2
pandas:               1.5.3
pyarrow:              15.0.1
pydantic:             2.6.3
pyiceberg:            <not installed>
pyxlsb:               <not installed>
sqlalchemy:           1.4.52
xlsx2csv:             <not installed>
xlsxwriter:           <not installed>

[cmdlineluser]

I think this is by-design:

These NaN values are considered to be a type of floating point data rather than missing data.

• https://docs.pola.rs/user-guide/expressions/missing-data/#notanumber-or-nan-values

[orlp]

This is by design. Use .fill_nan(None) first if you want to also fill NaNs.