(🎁) Narrow dict(`TypedDict`?) entry on item assignment

from typing import TypedDict

class Data(TypedDict):
    a: int | None
    
d1: Data
d1['a'] = 1
d1['a'] + 1  # error: None doesn't have +

d2: Data
d2['a'] = 1
assert isinstance(d2['a'], int)
d2['a'] + 1  # no error

d3: dict[str, int | None]
d3['a'] = 1
d3['a'] + 1  # error: None doesn't have +

d4: dict[str, int | None]
d4['a'] = 1
assert isinstance(d4['a'], int)
d4['a'] + 1  # no error

Is this intentional, in that assignment doesn't narrow because a custom dict subtype could modify __setitem__ to not set/set the item to something else? If so then the narrowing should apply to TypedDicts, because they are always vanilla dicts.

No implementation file or test is named in the issue. Start by reproducing the four TypedDict and dict assignment examples, then trace mypy's assignment-narrowing behavior; done means the intended TypedDict result is resolved consistently and covered by a regression test.