Hierarchical reindexing with NaN in index

[Marigold]

related #6322
dupe of #7031

I'm on 0.13.1, don't know if this is already fixed on master, but haven't found any issues about this. If I you have NaN value in MultiIndex, reindexing will fill the first value from reindexed series to all NaN indices. Perhaps an example will clarify it

Code:

import pandas as pd
import numpy as np

ix = pd.MultiIndex.from_tuples([('a',np.nan)], names=['col1', 'col2'])
pd.Series([1], index=[1]).reindex(ix, level=1)

Output:

col1  col2
a     NaN     1

Desired output:

col1  col2
a     NaN     NaN

[jreback]

dupe of #7031
in general this is an ambiguous operation and should be avoided (nan's in the multiindex)