FIX: resample to single group with custom function

[Komnomnomnom]

Using master code, when re-sampling to a frequency that would result in a single group the use of custom aggregation functions do not work as expected. The included changes fix this, and should fix #3849 too.

In [17]: import pandas as pd

In [18]: import numpy as np

In [19]: mysum = lambda x: x.sum()

In [20]: rng = pd.date_range('2000-1-1', '2000-1-10', freq='D')

In [21]: ts = pd.Series(np.random.randn(len(rng)), index=rng)

In [22]: ts.resample('M', how='sum')
Out[22]: 
2000-01-31    0.985236
Freq: M, dtype: float64

In [23]: ts.resample('M', how=mysum)
Out[23]: 
2000-01-31    0
Freq: M, dtype: float64

[jreback]

@jtratner take a look?

[jreback]

@Komnomnomnom can you add a release notes.....?

and add in tests for #3849
much appreciated

[Komnomnomnom]

Done and done (I presume you meant #3849 ... ?)

[jreback]

yep..that's what I meant!