Fix VirtualMetaclassType#implements? to ignore base_type

[straight-shoota]

This method override was plainly incorrect. Removing it defaults to the super implementation which is all that we want.

Fixes #12628

[HertzDevil]

It should be something like base_type.metaclass.implements?(other). Don't know if the CI crash is something else entirely.

[straight-shoota]

The failures point directly to one of the added specs, so it's definitely related.

Compiler specs succeeded on my machine, though 🤔

[straight-shoota]

One of the specs was buggy. It returned a metaclass instance from Program.run which apparently crashes (and that's probably fine). Just changed the spec code to make it work.

[HertzDevil]

Maybe even delegate the method to base_type.metaclass, since a similar delegate is present in Crystal::VirtualType.

[straight-shoota]

Do you have an example that fails without super || base_type.metaclass.implements?(other) (or delegating to base_type.metaclass)?

[HertzDevil]

Only base_type.metaclass.implements?(other) ensures T+.class is a subtype of T.class, which is consistent with how T+ is a subtype of T. Ideally both are false if we follow strict virtual type semantics, e.g. something like #10785, but unfortunately the compiler doesn't do that yet. So at the moment it is best to pursue consistency first. (The code class A; end; class B < A; end; x = A; pointerof(x).value = A || B will remain invalid.)