__r<magic>__ methods for reflected (swapped) operands doesnt duck type int to float

🐛 Bug Report

It seems like ints are not compatible with floats when used in the __r<magic>__ methods for reflected (swapped) operands like __radd__ the same was as in normal operands if the int/float is used as a type parameter in a generic type.

As a workaround it works if I specify a union type with the int case as well, but would be nice to not have to do that since its not needed otherwise.

To Reproduce

code.py:

from typing import List

class Vec:
    def __add__(self, other: List[float]) -> None:
        pass

    def __radd__(self, other: List[float]) -> None:
        pass

r1 = Vec() + [1]
r2 = [1] + Vec()  # error line
r3 = [1.0] + Vec()

Type checking the above code with python -m mypy code.py

Expected Behavior

I expect the r2 case to work as well

Actual Behavior

I get this error message for the r2 line: error: Unsupported operand types for + ("List[int]" and "Vec")

Your Environment

• Mypy version used: 0.782
• Mypy configuration options from mypy.ini (and other config files): No special config
• Python version used: CPython 3.8.4, 32 bit.
• Operating system and version: Windows 10

Start with the issue's code.py reproduction and run python -m mypy code.py using the reported cases. Trace how operand checking handles __add__ versus reflected __radd__ when List[int] is compared with List[float]; done means the r2 case type-checks without requiring a union workaround while existing cases remain correct.