[pydoclint] Ignore DOC201 when function name is "__new__"

[augustelalande]

Summary

Resolves #13079.

Do not enforce the documentation of return in a __new__ function. Arguably this could apply to some other dunder methods.

Test Plan

Added fixture.

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[AlexWaygood]

It is actually possible for __new__ methods to not return an instance of the class (e.g. https://github.com/python/cpython/blob/962304a54ca79da0838cf46dd4fb744045167cdd/Lib/pathlib/_local.py#L521-L524). But I agree that that's something of a rare edge case. We could in theory do some static analysis to figure out if __new__ returns an instance of the current class or not, but I think it's probably not worth it with our current type inference capabilities. So this LGTM.

I do also agree that we could probably extend it to some other dunders as well -- probably all of these?

ruff/crates/ruff_python_stdlib/src/typing.rs

Lines 355 to 370 in 7c872e6

	/// Returns the expected return type for a magic method.
	///
	/// See: <https://github.com/JelleZijlstra/autotyping/blob/0adba5ba0eee33c1de4ad9d0c79acfd737321dd9/autotyping/autotyping.py#L69-L91>
	pub fn simple_magic_return_type(method: &str) -> Option<&'static str> {
	match method {
	"__str__" | "__repr__" | "__format__" => Some("str"),
	"__bytes__" => Some("bytes"),
	"__len__" | "__length_hint__" | "__int__" | "__index__" => Some("int"),
	"__float__" => Some("float"),
	"__complex__" => Some("complex"),
	"__bool__" | "__contains__" | "__instancecheck__" | "__subclasscheck__" => Some("bool"),
	"__init__" | "__del__" | "__setattr__" | "__delattr__" | "__setitem__" | "__delitem__"
	| "__set__" => Some("None"),
	_ => None,
	}
	}

[AlexWaygood]

I do also agree that we could probably extend it to some other dunders as well -- probably all of these?

(These can be done as a followup, so I merged this for now :-)