Incorrect stub generation for subclasses of `pydantic.BaseModel`

Bug Report

stubgen does not properly generate ellipses for default values set in the scope of the class body of a Pydantic model.

To Reproduce

Running stubgen -m example

# example.py
from pydantic import BaseModel


class Foo(BaseModel):
    abc: int = 1
    xyz: str = "abc"

Should produce

Expected Behavior

# example.pyi
from pydantic import BaseModel

class Foo(BaseModel):
    abc: int = ...
    xyz: str = ...

stubgen should be populating the default annotation with ellipses, instead they are missing.

Actual Behavior

# example.pyi
from pydantic import BaseModel

class Foo(BaseModel):
    abc: int
    xyz: str

Your Environment

• Mypy version used: mypy 1.8.0 (compiled: yes)
• Mypy command-line flags: stubgen -m example
• Mypy configuration options from mypy.ini (and other config files): N/A
• Python version used: 3.11.8

Start by reproducing the issue with the shown example.py and stubgen -m example, then trace the stubgen path that handles annotated class-body defaults for subclasses of pydantic.BaseModel. Done means the generated example.pyi contains ... for both defaults, with regression coverage for this case.