ENH/API: add nlargest/nsmallest to DataFrame

[cpcloud]

xref: #7356

API like:

df.nlargest('column_name', 3)

impl like:

def nlargest(self, col, n):
    return self.loc[self[col].nlargest(n).index]

cc @hayd

[cpcloud]

@jreback i think actually you can do you something like and you can have a partial nested sort ...

def nlargest(self, columns, n):
    return self.loc[self[columns[0]].nlargest(n).index].sort(columns)

[cpcloud]

i'll move to 0.15 i think maybe mull over a bit