* Reorganise our sparql queries on the basis that Redland must be available, not only optional. So for anything querying the pool of data about plugins, we use a single datastore and model which is initialised at the outset by PluginRDFIndexer and then queried directly; for anything that "reads from a file" (e.g. loading annotations) we query directly using Rasqal, going to the datastore when we need additional plugin-related information. This may improve performance, but mostly it simplifies the code and fixes a serious issue with RDF import in the previous versions (namely that multiple sequential RDF imports would end up sharing the same RDF data pool!)