[Stubgen] Do not treat names ending in underscores as private

[inducer]

Right now, stubgen considers names ending in (single) underscores private;

nanobind/src/stubgen.py

Lines 788 to 799 in 109af19

	# Ignore private members unless the user requests their inclusion
	if (
	not self.include_private
	and name
	and not is_type_alias
	and len(name) > 2
	and (
	(name[0] == "_" and name[1] != "_")
	or (name[-1] == "_" and name[-2] != "_")
	)
	):
	return

This seems to go against common practice. For example, PEP8 states:

• _single_leading_underscore: weak “internal use” indicator. E.g. from M import * does not import objects whose name starts with an underscore.
• single_trailing_underscore_: used by convention to avoid conflicts with Python keyword [...]

It's easy enough to override in my own stub generation (override StubGen.put), but I do think it's poor default behavior. As a less-preferred option, one could make the private-name predicate customizable.

[wjakob]

Some Python builtins (e.g. enums) use a single underscore for implementation-specific bits, so having that policy as a default is unfortunately not negotiable. But I would be open to moving it into a separate method like self.is_private_name(name: str). (If you create a PR of this kind, please also include a brief docstring explaining the purpose and that the method is defined in this way to allow overriding in user code).