Do not treat annotated self-assign as special

[sterliakov]

Fixes #20144.

Historically we have special treatment for X = X "redefinitions" in global scope. This special case was introduced in #7144, and all test cases there are unannotated.

I suspect that it was not supposed to include annotated cases at all (cc @JukkaL as the PR author - is my understanding correct?), because it skips a whole bunch of annotation checks - final, classvar, annotation sanity, plugins, etc.:

mypy/mypy/semanal.py

Lines 3269 to 3280 in 12d5da4

	s.is_final_def = self.unwrap_final(s)
	self.analyze_lvalues(s)
	self.check_final_implicit_def(s)
	self.store_final_status(s)
	self.check_classvar(s)
	self.process_type_annotation(s)
	self.apply_dynamic_class_hook(s)
	if not s.type:
	self.process_module_assignment(s.lvalues, s.rvalue, s)
	self.process__all__(s)
	self.process__deletable__(s)
	self.process__slots__(s)

[github-actions]

According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅