It turned out that the costs of separating the XSD datatypes are too high
and probably not worth the effort, since with its limited scope
probably nobody would want to use XSD.ex outside of the RDF.ex context
anyway.
- we now only the store the lexical form when it's non-canonical
- Literal validation and canonicalization
- a general RDF.Datatype.Test.Case
- also tested datatype implementations for booleans, integers, string and langStrings
- use literal sigils in Inspect implementation of Literals when possible
