Data Explorer: More gracefully handle summary statistics for all-null arrays for pandas and polars

[wesm]

As described in #4307, for pandas we were raising a warning for all null / NA arrays when computing summary stats. The polars implementation turned out to have a different bug, so that is fixed and tested here also.

QA Notes

Follow example reported in #4307 for both pandas and polars (substitute all None values for polars)

[petetronic]

Confirmed that this removes the warning for Pandas

[petetronic]

With:

import pandas as pd
import polars as pl
import numpy as np

...for our treatment of Polars, should np.nan be considered a NA value? Right now it does not seem to be (and the summary stats show NAN):

pl_na = pl.DataFrame({"missing": pl.Series([np.nan] * 5, dtype=pl.Float64)})

Where as, using None is seen as an NA value with Polars (and avoids showing summary stats):

pl_na = pl.DataFrame({"missing": pl.Series([None] * 5, dtype=pl.Float64)})

For Pandas, note we consider either None or np.nan to be NA (and note summary stats are not shown too):

[petetronic]

I've logged an issue to track the question with polars with np.nan, and will land this change.