Add mapK to Tracer, SpanBuilder, etc.

[NthPortal]

Add mapK to Tracer, Tracer.Meta, SpanBuilder, SpanOps, Span, Span.Backend, and InstrumentMeta.

Add KindTransformer type used by mapK implementation, allowing for a single mapK implementation that supports many transformation target types (OptionT, EitherT, IorT, StateT, Kliesli, Resource) with very little work needed to add additional or custom ones.

see #261 for previous work, and #264 for a less complex but less powerful alternative

• scaladocs
• rebase -i --autosquash

[NthPortal]

I managed to write up a PoC implementation of KindTransformer that defines G in terms of F. however, it turns out that the added flexibility of being able to theoretically mapK over F ~> H is undermined by the inability to summon the needed context bounds of an arbitrary H (e.g. Functor). additionally, such a definition makes it difficult or impossible to use KindTransformer for types of G that are defined only indirectly in terms of F (whether via type aliases, type members or implicit transformations), thus making the PoC implementation less versatile. so it is going to the code graveyard.

but for posterity...

trait KindTransformer[F[_], G[_[_], _]] {
  type GoF[x] = G[F, x]

  val liftK: F ~> GoF

  def mapK[H[_], A](ga: GoF[A])(f: F ~> H): G[H, A]

  final def liftFunctionK[H[_]](f: F ~> H): GoF ~> G[H, *] =
    new GoF ~> G[H, *] {
      def apply[A](ga: GoF[A]): G[H, A] = mapK(ga)(f)
    }
}

[NthPortal]

rebased to fix conflict; autosquashed while I was at it

[NthPortal]

what's needed to move this PR forward? cc @iRevive @rossabaker

[iRevive]

Which laws KindTransformer should obey? Perhaps we can use Discipline to run law tests? https://typelevel.org/cats/typeclasses/lawtesting.html

[NthPortal]

I don't know and that sounds like a great idea!

[rossabaker]

I talked to @NthPortal today and we agreed laws are a good idea but shouldn't block progress.

I'll also search through the archives of my brain and try to think of Haskell precedent for similar, but this is good for now.

[iRevive]

@NthPortal apologies for making you wait for so long. The changes look good to me.
I have a few questions, but we can merge without addressing them.

[iRevive]

@rossabaker @armanbilge what's your opinion on these changes?

[rossabaker]

I talked to @NthPortal today and we agreed laws are a good idea but shouldn't block progress.

I'll also search through the archives of my brain and try to think of Haskell precedent for similar, but this is good for now.