[Mono] Enable support for IsByRefLike types in generics

[lambdageek]

Contributes to #65112

• [mono] Support AcceptByRefLike metadata generic param attribute #94421
• JIT+AOT boxing elision support
  • generic sharing support: As a first cut we should probably just always create specific instance code for byreflike instances. On the other hand, the gpAcceptByRefLike param may allow gsharedvt to do a better job (since we know we would never need to box values of the type). Done in [mono] Add basic ref struct support for generic parameter #99081
• Interp boxing elision support
  • [mono] box related IL instruction sequence optimization for interpreter #99379
• Need to delay type loading error until runtime for array of T with ByRefLike type #99165

The runtime design is specced out here: https://github.com/dotnet/runtime/blob/main/docs/design/features/byreflike-generics.md

There are tests in https://github.com/dotnet/runtime/tree/main/src/tests/Loader/classloader/generics/ByRefLike

[lambdageek]

/cc @SamMonoRT

[lambdageek]

We're gonna have a problem is mono_class_is_assignable_from_internal - we will need to make sure that we don't allow casting a ref struct RS to an interface I that it implements. The only way that kind of thing is allowed is if a generic param that allows ref structs is instantiated with a ref struct and then the method casts the generic param to the interface.

interface I {
   void M();
}
ref struct RS : I {
   public void M () { ... }
}

static class C {
   public static void BadClient (I i) {
      i.M(); // ok, it's an interface method call
   }
   public static void GoodClient<T> (T t) where T : allow ref struct, I {
      t.M(); // also ok, constrained.callvirt - an interface method call through a gparam
      
      // both of these are not ok. it's a "box !!0 ; callvirt"
      // I i = (I)t;
      // i.M(); // illegal
      // - or -
      // ((I)t).M(); // illegal
   }
}

RS s;
// I i1 = (I)s; // illegal - "box"
// C.BadClient (s); // illegal - also "box"
C.GoodClient<RS>(s); // okay, it's a call to a generic method instantiation.

[steveisok]

Removed #99820 from the list since it can be done separately / not at all.