Get Semigroupal from Arrow for free

[kailuowang]

We could add the following to Semigroupal's companion

  implicit def catsSemigroupalForArrow[F[_, _], A](implicit F: Arrow[F]): Semigroupal[F[A, ?]] = new Semigroupal[F[A, ?]] {
    def product[B, C](fb: F[A, B], fc: F[A, C]): F[A, (B, C)] =
      F.andThen(F.lift((x: A) => (x, x)), F.split(fb, fc))
  }

#2063 is adding a merge method, so this instance can just delegate to that.
To test this we might need to create a new case class MyFunc[A, B](f: A => B) and an Arrow instance for it and then test the Semigroupal laws against it.

I set the milestone to be 1.1, but it can certainly be done by 1.0

[LukaJCB]

Would this be a binary compatible change?

[kailuowang]

I think so.

[ceedubs]

A relevant question would be whether there are conceivable F types that form an Arrow but that we wouldn't want to derive this implicit Semigroupal for. I can't think of anything that might fall into that realm, but it's probably worth thinking about.

[kailuowang]

@ceedubs are you suggesting making it explicit?

[ceedubs]

@kailuowang that was one thought that I had. But I don't know if there's really any reason to that. Just my usual paranoia :). I'll ponder this a bit and give others a chance to weigh in over the next day or two, and if nobody can think of a reason that this implicit def shouldn't exist, then maybe it's either good or mostly harmless.

[stephen-lazaro]

I can open a PR for this. I think you can reasonably strengthen it to Apply as well, after playing with it a bit (sorry Kailuo, I mislead you the other day :) ). Given that that's a fair amount stronger, it might lend more weight to explicitness over implicitness here.

[LukaJCB]

Fixed by #2078