type DataFrame, Column, and GroupBy as Protocol

[MarcoGorelli]

technically we're defining protocols here - classes which must have certain attributes and methods

might as well type these as Protocol anyway. this will make type checking for implementations simpler, as it tells you whether anything's missing

minimal demo of how this helps:

(.venv) marcogorelli@DESKTOP-U8OKFP3:~/tmp$ cat t.py
from __future__ import annotations

from typing import Protocol
from typing_extensions import Self


class Cat(Protocol):
    def add(self, other: Self) -> Self:
        ...
    def sub(self, other: Self) -> Self:
        ...


class FancyCat(Cat):
    def __init__(self, value: int):
        self._value = value

    def add(self, other: Self) -> FancyCat:
        return FancyCat(self._value + other._value)

(.venv) marcogorelli@DESKTOP-U8OKFP3:~/tmp$ mypy t.py
t.py:19: error: Cannot instantiate abstract class "FancyCat" with abstract attribute "sub"  [abstract]
Found 1 error in 1 file (checked 1 source file)

"oh, right, I didn't implement sub. let me add that"

+    def sub(self, other: Self) -> FancyCat:
+        return FancyCat(self._value - other._value)

$ mypy t.py
Success: no issues found in 1 source file

[MarcoGorelli]

nobody's objecting, and type checking doesn't completely work properly without this, so let's merge