Ensure pyarrow dtypes

[felixschmitz]

One thing I noted when looking at the example data is that (some of?) the categoricals don't seem to catch the arrow-based missings, see here:

In [11]: x[["p_id_father", "birth_month"]]
Out[11]: 
          p_id_father birth_month
0                <NA>         NaN
1                <NA>         NaN
2                <NA>         NaN
3                <NA>         NaN
4                <NA>         NaN
...               ...         ...
29443775         <NA>           2
29443776         <NA>           7
29443777         <NA>           7
29443778         <NA>           7
29443779         <NA>           7

Please have a look at that, could be in a different PR though.

Originally posted by @hmgaudecker in #62 (review)

[hmgaudecker]

Changed the heading since the dtype relates to the entire column; it just happens to be visible in this view in how missings look like.