Columns with null dtype are ignored in sum_horizontal/mean_horizontal regardless of ignore_nulls parameter

[mcrumiller]

Issue description

If one of the columns in sum_horizontal or mean_horizontal has a null dtype then it is ignored even if ignore_nulls = False.

import polars as pl

df = pl.DataFrame({
    "a": [None, None],
    "b": [1, 2],
})
df.select(pl.sum_horizontal("a", "b", ignore_nulls=False))
shape: (2, 1)
# ┌─────┐
# │ a   │
# │ --- │
# │ i64 │
# ╞═════╡
# │ 1   │
# │ 2   │
# └─────┘

Installed versions

main

[ritchie46]

The above example is what I would expect. We don't propagate nulls in the sum, but use the identity.