DOC: document level-dropping for empty columns

[chris-b1]

I might have missed it, but I couldn't find this behavior documented anywhere. It is occasionally useful, if a bit magical.

In [1]: df1 = pd.DataFrame([1,2,3], columns=[['a'],['b']])

In [2]: df2 = pd.DataFrame([1,2,3], columns=[['a'],['']])

In [5]: df1['a'], type(df1['a'])
Out[5]: 
(   b
 0  1
 1  2
 2  3, pandas.core.frame.DataFrame)

In [6]: df2['a'], type(df2['a'])
Out[6]: 
(0    1
 1    2
 2    3
 Name: a, dtype: int64, pandas.core.series.Series)

[jreback]

yeh '' is an empty level for dropping purposes. The empty level concept is much better now-adays IOW NaN is pretty good. So not really sure this is necessary any longer.

All of that said. This is pretty buried in pretty tangled code.