Inappropriate result of expr skew and kurtosis for a magic constant inputs

[j7168908jx]

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

I encountered this inappropriate output at this very exact float number:

import polars as pl
import scipy.stats as st
import pandas as pd

a = pl.DataFrame({"a": [1.0042855193121334] * 11})
b = pd.DataFrame({"a": [1.0042855193121334] * 11})

print(a.select(pl.col('a').kurtosis()).item())
print(b.a.kurtosis())
print(st.kurtosis([1.0042855193121334] * 11))


print(a.select(pl.col('a').skew()).item())
print(b.a.skew())
print(st.skew([1.0042855193121334] * 11))

and the output is

-2.0
0.0
nan
-1.0
0.0
nan

where, the appropriate output should be either 0.0 or nan.

Changing this floating number to others will cause the normal nan output.

>>> print(pl.DataFrame({"a": [1.0042834] * 11}).select(pl.col('a').kurtosis()).item())
nan

Log output

I did not obtain log here.

Issue description

At this very exact number 1.0042855193121334 replicate exact 11 times, the output of polars' skew and kurtosis seems unreasonable. (no matter setting which value for bias or fisher in skew or kurtosis's parameter)

Expected behavior

The appropriate output should be either 0.0 or nan.

Installed versions

--------Version info---------
Polars:               0.20.15
Index type:           UInt32
Platform:             Linux-5.19.0-1010-nvidia-lowlatency-x86_64-with-glibc2.35
Python:               3.10.13 (main, Feb 27 2024, 12:25:06) [GCC 11.4.0]

----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
numpy:                1.26.4
openpyxl:             <not installed>
pandas:               2.2.1
pyarrow:              15.0.0
pydantic:             <not installed>
pyiceberg:            <not installed>
pyxlsb:               <not installed>
sqlalchemy:           <not installed>
xlsx2csv:             <not installed>
xlsxwriter:           <not installed>

[deanm0000]

This should be fixed with #15137