Preserve Class<T> type arguments

[RushikeshGandhmal]

Fixes #1729

Previously, Class<T> was explicitly erased to Class during type resolution, so C is Class<A> could not be distinguished from C is Class<B>.

This change introduces a ClassValueTypeNode that:

• preserves the type argument through type resolution and runtime checks
• validates class values by requiring the represented class to be a subclass of the type argument
• rejects invalid concrete type arguments such as unions, nullables, constrained types, parameterized types, and modules
• still allows type variables like Class<Type> used in stdlib generic signatures

A regression test is added covering:

• positive cases (C is Class<A> vs Class<B>)
• aliases to class types and alias-preserving generics
• exported schema preservation
• stdlib compatibility with type variables
• rejection of invalid Class<T> arguments

[RushikeshGandhmal]

Hi @bioball -- this is a more substantial type-system fix for #1729. I kept the change focused on Class resolution and used the existing ValidatingObjectSlotTypeNode pattern for validation. The regression test covers both the new semantics and the stdlib compatibility case. Thanks for taking a look!

[HT154]

This appears to be a duplicate of #1698. If you have input on this change, I would encourage you to discuss there so we can avoid splitting the discussion.

[RushikeshGandhmal]

Thanks for pointing this out. I missed #1698, and after comparing the implementations I agree that #1763 overlaps it. I’ll close this PR and move the additional schema-export observation and regression test to #1698 so the discussion stays in one place.