Fix Ref<>.is_valid() for ScriptInstanceExtension

[Gallilus]

First, check the reference then check the value.

Fix #78116
caused by #63712
#63712 (comment)

[Gallilus]

I still need to do some functional tests do double-check.

[AThousandShips]

If you need time to work more you can convert it to a draft PR, but shouldn't be needed, assigned to 4.1 tentatively as it is a trivial fix that I can't see causing any regressions

[Gallilus]

It is an improvement! Not jet sure it is a catch-all.

[Gallilus]

Using reference_2b I am able to get the following access error.
So I will change the PR to if (script.is_valid() && pcount > 0) { it seems to work.
Will test it in the next few days during the cleanup of my code.

    MyClass my_object = memnew(MyClass);
    Ref<MyClass > reference_1 = *my_object;
    Ref<MyClass > reference_2a = reference_1;
    Ref<MyClass > reference_2b = reference_1.ptr();
    if (refrence_2b.is_valid()){
        // Is reference_2 a or b valid?
        // Will reference.is_valid() && refrence->is_valid() catch this problem?
    }

[YuriSizov]

The consensus seems to be that the change is correct, and it looks like that to me as well.

[YuriSizov]

Thanks!