How we handle @final decorator in the stubs

[loicdiridollou]

Opening this issue more as a discussion, I was doing some cleanup with the stubtest task and I have noticed that some of the methods in pandas annotated with @final are not always annotated as such in the stubs.
This generates around 500 warnings with the stubtest tasks. I know we have discussed it some times in PRs but wanted to open this issue to clarify how we should stand on it.
The main points are:

• Should we also annotate the stubs with @final (it is partially done right now)?
• What is the impact of having it? (or not having it)
• Does it make a difference to have it or not?

[Dr-Irv]

• Should we also annotate the stubs with @final (it is partially done right now)?

Yes.

• What is the impact of having it? (or not having it)

By having it, we are signaling that the methods cannot be overridden in user code.

• Does it make a difference to have it or not?

See above - provides better type checking in case people try to override our methods.