adapter: Refactor XML and JSON de-/serialization methods

[zrgt]

The methods for handling attributes of abstract classes seemed to me too big.

In this PR these methods are refactored. Mostly a method extraction refactoring was done, to simplify the methods and to keep less abstraction levels in a single method

The function _expect_type was renamed to _is_of_type as it represents clearer what the func does.

This PR is not relevant for the V3.0 release and can be merged later

[jkhsjdhjs]

I don't think it's necessary to create a new function for amending HasExtension/HasSemantics/etc. attributes in each (de-)serialization, as they only have one usage each and thus don't aid in reducing duplicate code. I wouldn't mind them for the purpose of shortening the other function though, but that wasn't too long either IMO.
But while you're at it, you can now also put the _amend_*_attrs() functions in a dict with the corresponding type and perform the isinstance() checks in a loop over said dict.

Maybe you can also unify the naming of the amend_*_attributes and extend_with_*_attrs functions to amend_*_attributes, which fits best IMO.

[zrgt]

I don't think it's necessary to create a new function for amending HasExtension/HasSemantics/etc. attributes in each (de-)serialization, as they only have one usage each and thus don't aid in reducing duplicate code. I wouldn't mind them for the purpose of shortening the other function though, but that wasn't too long either IMO.

The purpose of the refactoring is not only reducing duplicated code, but also improving readability.
The function was 50 lines long, which IMO too long. The function had multiple levels of abstraction before, so the readability and further development was affected.

[zrgt]

But while you're at it, you can now also put the _amend_*_attrs() functions in a dict with the corresponding type and perform the isinstance() checks in a loop over said dict.

Good Idea! But may any problems occure because of the order?

[zrgt]

Maybe you can also unify the naming of the amend_*_attributes and extend_with_*_attrs functions to amend_*_attributes, which fits best IMO.

extend_with_*_attrs is used in serialization while in deserialization amend_with_*_attrs is used

[s-heppner]

I believe this refactoring made the code more readable and I don't have any strong opinions on the remaining details, so I let you two work them out.

[s-heppner]

Let's discuss the super.__init__() removal, but other than that it LGTM