Groupby doesn't call aggregation on empty groups

[TomAugspurger]

This should exit the interpreter.

In [1]: import pandas as pd

In [2]: df = pd.DataFrame({"A": pd.Categorical(['a', 'a'], categories=['a', 'b']),
   ...:                    "B": [1, 1]})
   ...:

In [3]: import sys

In [4]: def f(x):
   ...:     if len(x) == 0:
   ...:         sys.exit(1)
   ...:     return len(x)
   ...:

In [5]: df.groupby('A').agg(f)
Out[5]:
     B
A
a  2.0
b  NaN

Instead, I think groupby assumes the output of the custom aggfunc is NaN.

[arw2019]

This is still a problem on 1.2 master

[rhshadrach]

This is fixed on main, could use tests.