Don't flag match type aliases as unreducible

[odersky]

If M[_] is a match type alias, it was rejected for being an unreducible
application of wildcard types. It is now accepted. I believe that is sound sinxe
it is simply an unreducible match type. The situation is not the same
as F[_] where F is an abstract type that can be refined in several ways
in subclasses.

Fixes #9999
Fixes #5592

[smarter]

The situation is not the same as F[_] where F is an abstract type that can be refined in several ways in subclasses.

I don't think it's fundamentally different since a match type application can reduce to a regular type application:

class A {
  type F[X]
  type M[X] = X match {
    case Int => F[X]
    case _   => F[X]
  }

  def foo(x: M[_]): M[_] = x
}

Though my recollection is that allowing unrestricted wildcard applications is not by itself unsound, the problems come from trying to decide which types are subtype of that type and getting it wrong, but if we're conservative and accept that M[_] has no known subtype, then I believe we're OK.

[odersky]

@smarter Yes, in your example M[_] would not reduce to F[_].