Add context to annotation covered text

The full annotation context should be composed of:

• context left
• original (actual) selection
• context right

This triple should be unique within the entire document, at the time of creation. A simple method of ensuring uniqueness is to start from the original selection and incrementally wrap it with surrounding text until the triple becomes unique. If the original selection is already unique within the document, no additional context should be required.

This context should come as an addition to the already existing offset-based context and should be used when the offset determined section in the current text does not match the stored original selection. In this case, the left and right context should be used to try to redetermine the annotated section within the modified document.

So it applies to both modified documents and also to annotators that currently run on a converted version of a document and want to use feedback data that has offsets for the original version of the document.

This should probably be done in the CAS Consumer, if the annotator did not already do that.