feat: Add mean_horizontal for temporal dtypes

[mcrumiller]

Closes #20313. Adds a path in mean_horizontal for temporals. A bit of logic update:

1. Checks to ensure that if we have numeric columns, all columns are numeric (includes bool/null).
2. Checks to ensure that if we have a temporal column, all columns have the same temporal dtype.

This implementation doesn't allow taking the mean of mixed time-unit dtypes that are otherwise the same. I am not sure if I should update logic to include this or not. In other words:

from datetime import datetime
import polars as pl

dates = [datetime(2025, 1, 1), datetime(2025, 1, 2)]

df = pl.DataFrame({
    "ms": pl.Series(dates, dtype=pl.Datetime(time_unit="ms")),
    "us1": pl.Series(dates, dtype=pl.Datetime(time_unit="us")),
    "us2": pl.Series(dates, dtype=pl.Datetime(time_unit="us")),
})

df.with_columns(
    pl.mean_horizontal("ms", "us1"),   # error--dtypes are not compatible
    pl.mean_horizontal("us1", "us2"),  # ok
)