DataFrame.Groupby.Rolling is missing proper typing

[Dr-Irv]

Describe the bug

In this line:

pandas-stubs/pandas-stubs/core/groupby/groupby.pyi

Line 64 in 6fe90bb

def rolling(self, *args, **kwargs): ...

we are not returning a Rolling object with the proper subtype of Series or DataFrame

To Reproduce

from typing_extensions import reveal_type
import pandas as pd

df = pd.DataFrame(
    {"A": [1, 1, 2, 2], "B": [1, 2, 3, 4], "C": [0.362, 0.227, 1.267, -0.562]}
)

d1 = df.groupby("A").rolling(2).sum()

s1 = df.groupby("A")["B"].rolling(2).sum()

s2 = df.groupby("A").rolling(2)["B"].sum()

reveal_type(d1)

reveal_type(s1)

reveal_type(s2)

at runtime produces

Runtime type is 'DataFrame'
Runtime type is 'Series'
Runtime type is 'Series'

but pyright produces

  grbroll.py:14:13 - information: Type of "d1" is "Unknown"
  grbroll.py:16:13 - information: Type of "s1" is "Unknown"
  grbroll.py:18:13 - information: Type of "s2" is "Unknown"

Please complete the following information:

• OS: Windows 10
• python version: 3.9
• version of type checker: pyright 1.1.313
• version of installed pandas-stubs : 2.0.2.230605

Additional context

This will need some care to make sure that the types are done correctly. May have to move the def rolling from pandas-stubs/core/groupby/groupby.pyi to pandas-stubs/core/groupby/generic.pyi in DataFrameGroupby and SeriesGroupBy.