Treat XHTML at a primitive FHIR type, not a primitive.

[ewoutkramer]

In our POCO's the Narrative.Div is a special case. The FHIR spec defines it as a normal element of type XHtml (and we have generated that type), but we have represented that element as a string. Consequently, both ITypedElement and IReadOnlyDictionary have to jump through hoops to make this work like the specification. It is probably better to consider XHtml to be a primitive FHIR type for which the XML serialization has no capabilities for presenting the extensions or id (it could be done in json!).

This is less breaking than it sounds, since in this case, the existing Narrative.Div element remains as a convenience property (just like with all primitives) of type string, we'll be adding a DivElement property, and move the attributes over. This might even allow us to generate the Narrative type, since those are currently hand-written because of this exceptional situation.