Consider creating a UninferableType or _Uninferable class

[DanielNoord]

I opened microsoft/pyright#3641 as I wondered why pyright didn't recognise how we type Uninferable. Normally they are a little bit more up to date than mypy so I wondered if this was intentional.

Turns out it is. According to them, the way we currently handle the typing of Uninferable is incorrect and should ideally be refactored.
As we're stille early days into the typing of astroid I think there is still chance to do this.

Their suggestion is to create a private class (_Uninferable) which Uninferable can then instantiate. One of the issues with this is that we tend to require Uninferable as a type in pylint as well and so we would need to import that private class as well.
We could also create a public class, perhaps suffixed with Type, which we document as only being useful for typing.

Let me know if you guys thinks this is something we should do and what approach is best.

/CC @cdce8p As you're likely interested in this.

[DanielNoord]

@jacobtylerwalls Do you have a preference for either of the two solutions?

[jacobtylerwalls]

I think UninferableType is more clear. I wish PEP 661 had an implementation.