mask values by isnan for dataframe groupby with dropna=True

[stress-tess]

In [32]: df = ak.DataFrame({"A":[1,2,2,np.nan],"B":[3,4,5,6]})

In [33]: df.groupby("A").count()
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)

ValueError: Attempt to group array using key array of different length

>>> df.groupby("A",dropna=False).count()
     B
A
1.0  1
2.0  2
NaN  1 (3 rows x 1 columns)

doing a dataframe groupby on a column containing NaNs causes an error because we are trying do the aggregation on the full column length. We need to be masking out the values in the NaN segment

>>> df.to_pandas().groupby("A").count()
     B
A     
1.0  1
2.0  2