Power to the NA with Int64 series gives NA

[jorisvandenbossche]

Corner-case, but the integer NA type does seem to behave differently when it comes to rpow:

In [57]: pd.Series([1, 1, 1]) ** pd.Series([1, np.nan, 3])
Out[57]: 
0    1.0
1    1.0
2    1.0
dtype: float64

In [58]: pd.Series([1, 1, 1]) ** pd.Series([1, np.nan, 3], dtype='Int64')
Out[58]: 
0      1
1    NaN
2      1
dtype: Int64

In [63]: 1 ** np.nan
Out[63]: 1.0

[TomAugspurger]

Closely related, object gets this wrong too.

In [6]: s = pd.Series([1, None], dtype=object)

In [7]: 1 ** s
Out[7]:
0      1
1    NaN
dtype: object