Improve Tuple and NamedTuple

[odersky]

Tuple is a very important core class of the Scala 3 standard library. But its organization was a jumble of different styles, with major omissions in functionality and a lack of consistency.

I made an effort in #19172 to improve Tuple.scala and in #19174 to align NamedTuple.scala and Tuple.scala. The work is not finished yet, and it would be great if someone could continue it. In particular:

• Make sure that every type we define has a corresponding method implementing that type, or if that does not make sense, explain in the doc comment of the type why there is no method.
• Make sure doc comments are consistent and helpful.
• Make sure we have tests for all operations.
• Have we missed important operations?
• Make sure Tuple and NamedTuple are aligned.

One area where NamedTuple and Tuple are not aligned is that there is no concept of "non-empty tuple" for named tuples. I believe adding that to Tuple was a big mistake. If you look at the definition of Tuple you'll find seemingly arbitrary and non-sensical definitions that sometimes use Tuple and sometimes use EmptyTuple. For instance

  type Head[X <: NonEmptyTuple]
  type Last[X <: Tuple]

Why is Head only defined for NonEmptyTuple whereas Last is defined for Tuple? That makes no sense!

The problem comes up because "being non empty" is a structural property of a tuple and we want to express it with a nominal type. That's usually a bad idea since nominal types are just not expressive enough. So the end result is that we shoot ourselves in the foot, and that for zero gains in safety. In fact, Last above is just as safe as Head. When applied to the empty tuple, neither of them reduces to a type.

[nicolasstucki]

We have a similar issue with NonEmptyTuple.apply. It should be defined in Tuple.

[EugeneFlesselle]

There have a few been attempts at improving the generic tuples, which have proven hard to get merged.
I am splitting the proposed changes in different PRs, the list of which I will keep updated below:

• Move NonEmptyTuple members into Tuple #21291
  • based on Move NamedTuple.head to NamedTupleDecomposition #21308
• Refine the bounds of the Tuple.Filter type lambda predicate .. #21286
• Add an infix shorthand for Tuple.{Append, Concat} #21288
• Combine cases of Tuple.Zip disjoint from (h1 *: t1, h2 *: t2) #21287

Which of those are accepted should be decided and discussed by the core team.