RFC: Add DelegateFactory and DelegateInvocationHandler

[rschmitt]

At a minimum this needs to be gussied up with Javadoc, but I want to ensure that the feature is thought through conceptually.

[rschmitt]

Issues so far with this code:

• Methods on java.lang.Object are indistinguishable from @Override methods on the proxy class itself
• Default methods on a superinterface are also indistinguishable from @Override methods on the proxy class itself

In both of these cases, the workaround would be to define an explicit delegation method, which is the opposite of what we actually want. The other workaround would be to ignore the superMethod and just delegate everything to the delegate unless an override can be found directly on the proxy class. However, I think what we actually want here is a more general interface. Instead of just passing in a MethodHandle superMethod, we want a tuple of (superMethod, source), where source can be an enum (OBJECT, DEFAULT, BASE_CLASS) or a Class or something.

[rschmitt]

This appears to be working so far:

        if (superMethod != null)
            try {
                // Before binding to superMethod, we have to verify that we're not actually binding
                // to a method supplied by java.lang.Object, or a default superinterface method.
                delegateClass.getDeclaredMethod(methodName, methodType.dropParameterTypes(0, 1).parameterArray());
                return new ConstantCallSite(superMethod.asType(methodType));
            } catch (NoSuchMethodException ignore) { }

[bdonlan]

Using that method to probe for supers does mean that this may have surprising behavior when you subclass these wrappers (which is, admittedly, somewhat silly). Maybe it'd be better to just check for methods that are implemented in Object (or an interface default) vs other classes in the super chain.

I'd also prefer to have the delegate be provided by a getDelegate() or similar method; this allows for e.g. wrappers that check that a delegate hasn't been invalidated on every call.