Implement annotator feedback mechanism

The knowledgebase already contains feedback information that is currently not exposed to annotators that might be capable of learning.

The proposed solution is, within an analysis process, to flood the CAS with annotations for the current document that is being processed.

This way, each annotator could have access to all previous annotations, each of them with their corresponding validation status and named entities.

The big advantage of this solution is that it is UIMA friendly and compatible with any language the annotators might be written in.
Of course, this also requires that the CAStypes reflect as closely as possible the model types.