Make Named Tuples a stable feature in 3.7

[WojciechMazur]

We make the Named Tuples again a standard feature, after they were delayed before 3.6 release #22045

Includes fixes to make deprecated syntax warnings and its migration still work - in many cases untpd.Tuple has become untpd.Parens

[Ichoran]

It would be helpful to have, somewhere (ideally in a link from the PR that re-establishes it as not-experimental), a link to significant uses of named tuples. I've tried using them and it's not clear to me that they're pulling their weight at least with the loads I'm trying to haul. But others may have had a more appropriate use-case.

If the only bar is "this isn't so obviously broken as to be useless", the bar is well-cleared by now. But if the desired bar is "the advantages of the feature are worth the complexity", it would be nice to have it better documented that this is the case. (Of course it's something of a chicken-and-egg situation because people don't use experimental features as much in case they're dropped.)

[bishabosha]

@Ichoran I will unprivate before my Scalar talk, but I've prepared a webapp using named tuples for typed dataframe, server endpoints, json serialisation, ORM.
I'd like to push for wrapping over foreign memory api but not had time yet.
Also I wanted to do a config parser but I think the concept needs a redesign due to the lack of possibility for recursion.

[Ichoran]

I will be interested to see the typed dataframe implementation! That was one of my use cases too, and I wasn't happy with the ergonomics and safety of same-typed differently-named columns among other things. But I may not have chosen the best design.

[odersky]

LGTM