Allow bare typing.ClassVar qualifiers

[Viicos]

As per https://discuss.python.org/t/81705.

[carljm]

The substance of this looks good to me, thanks for the PR!

[carljm]

The listed options are not the only choices here, I'd prefer not to imply that they are. (I think a strong case can be made that inferring Any | Literal[2] is the most correct option, from a gradual-typing perspective.) I'm fine with just listing the simpler options that you have here, but we shouldn't imply that it's an exhaustive list.

Suggested change

	inferred as some type to which this value is :term:`assignable` (in this
	case, either ``int``, ``Literal[2]``, or ``Any``).
	inferred as some type to which this value is :term:`assignable` (for example,
	``int``, ``Literal[2]``, or ``Any``).

[Viicos]

This is what I wanted to write out to begin with, but I realize my wording is wrong indeed :)

[carljm]

The only dynamic type discussed in the typing spec is Any, so it's a bit odd to imply here that there would be others. Most type checkers do distinguish between different "origins" of Any, and some do so with a variant of Any named Unknown, but as far as the spec is concerned these are all just Any.

Suggested change

	should be inferred to an unknown static type (such as :ref:`Any`).
	should be inferred as :ref:`Any`.

[Viicos]

I was wondering if such a concept of an unknown static type, as written in the referenced any section:

typing/docs/spec/special-types.rst

Lines 6 to 14 in fafcdeb

	.. _`any`:
	``Any``
	-------
	``Any`` represents an unknown static type.
	Every type is :term:`assignable` to ``Any``, and ``Any`` is assignable to every
	type.

would be worth specifying, and that's why I chose the same phrasing. But let's not complicate things, I'll go with Any.

[erictraut]

This change to the spec looks reasonable to me. If this is accepted, someone will need to write conformance tests for it. @Viicos, if you're willing, you could include the conformance test updates in the same PR.

[Viicos]

I created the TC request: python/typing-council#43.

[rchen152]

Looks good to me! For what it's worth, a bare ClassVar with no assigned value looks a little fishy to me, but I suppose if x: ClassVar[int] with no value is allowed, allowing x: ClassVar makes sense.

[JelleZijlstra]

All approved now, thanks Jukka!

@Viicos would you mind regenerating version.toml etcetera? Unfortunately those are prone to meaningless merge conflicts.