Relaxing 3 layer (source, dataset, version) naming

Note: these are early notes; this is not currently implemented. Vote for #174 if you need this.

When a source adopts your enhancement params, they don't want your source identifier in their dataset URIs.

the source/SOURCE-ID step in the URI construction can already be omitted for the data-level triples using conversion:domain_template, conversion:equivalent_property, and conversion:range_template, but not for the dataset URIs that contain those data-level triples. To omit source/SOURCE-ID, assert the following in your enhancement parameters:

   conversion:source_identifier conversion:null;

This is a new use for conversion:null, which used to be used only to say that certain triples should not be asserted. This same meaning is being reapplied to say that no source identifier should be asserted.

This is very different from not specifying a value, or specifying an empty value (BOTH of these are an ERROR):

#   conversion:source_identifier ""; # no source-identifier every asserted is an error.

   conversion:source_identifier ""; # an empty source_identifier is an error.

In either of the cases above, the following RDF will be the only output the converter provides:

_:node166lm2ec8x1 
   rdfs:comment "Error: source_identifier, dataset_identifier, and version_identifier must be specified for all conversions." ;
   rdfs:seeAlso <https://github.com/timrdf/csv2rdf4lod-automation/wiki/Relaxing-3-layer-%28source,-dataset,-version%29-naming> , 
                <https://github.com/timrdf/csv2rdf4lod-automation/wiki/Conversion-process-phase:-name> .