Fix isSubType(A#X, B#X) not checking if A <: B

[smarter]

Review by @odersky @namin

[smarter]

On a related not, the following typechecks with scalac but not dotc, I don't know if this is intentional:
https://gist.github.com/anonymous/bce9371ffaa2e7edaaa9

[odersky]

The status of A#X wrt subtpying is unclear. "A <: B implies A # X <: B # X "is not supported by DOT. Do without further study we should not add it, because it might well be unsound.

[samuelgruetter]

The DOT calculus does not even have types of the form A#X (yet), but only of the form p.X, where p is a stable path.
But as far as I understand, this PR adds more checks to isSubType, I think it's "safer" to merge it than not to merge it.

[namin]

In any case, we ought to document those tricky cases that we should investigate outside the pure calculus.

[smarter]

Okay, so does that mean that this comment: https://github.com/lampepfl/dotty/blob/f57753ff673d39ab68f2182459d5cfb2125d15a1/src/dotty/tools/dotc/core/TypeComparer.scala#L466-470 is wrong?

[odersky]

The particular PR is good, I think. We need to come back to the status of # projection in DOT. I'd prefer to do completely without # at the moment, but unfortunately the higher-kinded types encoding requires unrestricted #, so our choice is either no higher-kinded types or an operator which we do not know that well. But that's an aside. The PR itself is good.

[odersky]

Thanks for the contribution, Guillaume!