Increase laziness of LazyList.cons.apply

[NthPortal]

Change LazyList.cons.apply so that the tail computation
is wrapped as a state computation for a new LazyList,
thus deferring its execution as long as possible.

Follow-up to #8985

---

Before (Scala 2.13.3-):

scala 2.13.3> LazyList.cons(1, { println("hey"); LazyList() })
val res6: scala.collection.immutable.LazyList[Int] = LazyList(<not computed>)

scala 2.13.3> res6.head
hey
val res7: Int = 1

After (Scala 2.13.4+):

scala 2.13.4> LazyList.cons(1, { println("hey"); LazyList() })
val res6: scala.collection.immutable.LazyList[Int] = LazyList(<not computed>)

scala 2.13.4> res6.head
val res7: Int = 1

[NthPortal]

I noticed there were no tests for these while I was in the area, so I added tests

[SethTisue]

@NthPortal would you mind improving the PR title and description to be at least a little more informative for end users? since we're linking to this from the 2.13.4 release notes

perhaps you can borrow some verbiage from the discussion on #8985

[NthPortal]

@SethTisue is that okay?

[SethTisue]

@NthPortal better — if you want to improve it further, I would suggest adding an example code snippet that is affected by the change, and describe how it is affected. even though I've looked at this stuff multiple times before, when I see LazyList.cons.apply my reaction is still "huh? what's that?" and I have to go look it up. an example would really help, I think.

[som-snytt]

"If you're as lazy as LazyList, you probably don't need to know more."

Sometimes it helps to be usefully vague.

Do I need to learn this?

• No.