Overload validation does not properly handle implied subtype relationships (e.g. int and float)

I'm working to make pyright's overload validation work consistently with mypy's.

Mypy's overload validation behavior is spec'ed here: https://github.com/python/typing/issues/253#issuecomment-389262904

As part of this validation, mypy reports errors when there is overlap between two overloads that return different return types. The following test case shows four such examples. Mypy properly reports the error in three of the four cases, but it misses the case where there is an implied subtype relationship, as between int and float.

# pyright: strict

from typing import Literal, Union, overload

class Parent: ...
class Child(Parent): ...

# Test 1: Literal subtype
@overload
def foo1(x: Literal[3]) -> int: ...
@overload
def foo1(x: int) -> str: ...

# Test 2: Subclass subtype
@overload
def foo2(x: Child) -> str: ...
@overload
def foo2(x: Parent) -> int: ...

# Test 3: Implicit subtype
@overload
def foo3(x: int) -> str: ...  # Mypy does not report error here
@overload
def foo3(x: float) -> int: ...

# Test 4: Union subtype
@overload
def foo4(x: int) -> str: ...
@overload
def foo4(x: Union[int, str]) -> int: ...

まず、issue にある4つの overload の例を再現し、mypy の診断をリンク先の typing specification と比較します。次に、サブタイプの重複チェックを担当する overload 検証パスを追跡し、そのうえで int/float のケースが literal、subclass、union のケースと一貫して診断されることを確認します。