RDF standarization of vector datatype and toolkit

[kysrpex]

The current implementation of vectors is custom and does not adhere to any particular standard. This is specially troublesome when dealing with wrappers and RDF import/export, as there is no standard way to represent the exported data, and in the end each backend ends up with a different representation.

A better approach might be to define some kind of object vector, and one or more standards to represent it using linked data, e.g. RDF List (the most simple one) an EMMO-based representation, etc. The vector object would act as an easy to use interface based on something like numpy or pandas to the vectors, and maybe have a writeable property defining its internal representation. The corresponding triples would be stored in the in-memory graph, and thus imported/exported/moved transparently to the wrappers.

We may even consider going more general and implement tensor support instead.