Initial refactorings for the highlighting endpoint

Eventually fixes #529 (closed).

This MR includes changes to the pos-tagging code to make obvious which are the transformations happening on the tokens and how GGTX is transforming them to yield eventually an Ngrams type.

It also introduces a basic /documents/:docid/highlight endpoint (which is still a work in progress) to make it possible to highlight and normalise user-selected terms directly on the backend.