adapter.json.json_deserialization: change use of Enum-Dicts from adapter.json.json_serialization to adapter._generic

The functions still have the parameter "namespace", that changes the
namespace of the element they return, but if the function serializes
everything inside the element after an xsd-schema, it only makes sense
that every child has the namespace fixed to the one defined in the
schema and not to the one, that the outer most element might have.

Since the Dicts to serialize Enum classes are the same for both XML and
JSON, move them into a _generic module to avoid duplicate definition

# Conflicts:
#   requirements.txt

XSD-Schema demands element "<value>", even if it is empty

somehow, the option of the submodelElement to be a ReferenceElement was
missing

the element "<annotations>" is mandatory according to the xsd-schema,
even, if there are no references inside

since annotatedRelationshipElement inherits from RelationshipElement, an
object of class annotatedRelationshipElement is obviously also instance
of RelationshipElement, therefore, it would be wrong to serialize it
twice, once as annotatedRelationshipElement and once as
RelationshipElement, Fix by checking first for relationshipElement and
then inside that check for annotatedRelationshipElement.

as they do not exist in the schema

- add "<allowDuplicates>" child with text="false" (default value), since
this attribute does not exist in the pyass implementation

- change order of the <ordered> and <value> children, to fit the order
in the schema

According to the schema, child has to exist even if it is empty

According to the schema, the child has to exist, even if it is empty

According to the xsd-schema, the child <containedElements> has to exist
even if it is empty

change location of "kind" in the serialization to the end of the asset
element

the order/sequence of the serialization of attributes was wrong.
switched the place of "derivedFrom" and "assetRef" serialization