Context-aware indexing

[crtxcr]

Currently, text is text. There is no context information.

It would be great if we could to something like: "c:(word) type:heading" to search in headings of documents.

For PDFs, we may simply extract the index to do that, at least, as a first step.