Make horizontal and vertical versions of and(), any() consistent w.r.t default null strategy

[stfdxv]

Description

import polars as pl

df = pl.DataFrame({"a": [True, None], "b": [None, False]})

# Horizontal version uses Kleene logic by default
df_any_and_h = df.with_columns(
    any_h=pl.any_horizontal("a", "b"),
    all_h=pl.all_horizontal("a", "b"),
)
print(df_any_and_h)

# Vertical version ignores nulls by default
df_any_v = df.select(pl.any("a", "b"))
print(df_any_v)
df_all_v_default = df.select(pl.all("a", "b"))
print(df_all_v_default)

The default (and only) null strategy for horizontal version of and(), any() is to use Kleene logic. As if null represents boolean unknown.
The default null strategy for horizontal version of and(), any() is to ignore nulls. But can be changed to Kleene logic with ignore_nulls=False.
I expect them both to have the same default null strategy.

[coastalwhite]

This is basically already added with bitwise aggregations #18994