Attribute::fire_change_event: Create AttributeValue_4 for IDLv4 device

This change is "obvious" in the sense that without this change the vers == 4 branch cannot be taken. However, I cannot seem to write a test which demonstrates a problem when we don't have this change.

The only difference between AttributeValue_5 and AttributeValue_4 is the presence of a data_type member for AttributeValue_5. This seems to allow overriding the data_type that is sent to the subscriber (from the value in the AttrValUnion), but it is not clear to me when this is needed.

This suggests there might be some refactoring we can do to simplify things, but I am not brave enough to do this without more test coverage. I propose that we just merge this anyway as it is "obvious" and we can think about refactoring once we have improved the test coverage.

Closes !1226 (closed) and #1229.