Add a really basic prototype of MuPDF-based backend (for PDF documents) for Okular.

What it currently does is: loading, rendering, text extraction, and few document information.
A small Qt-only layer is done to encapsulate the fitz/mupdf API.

The development of MuPDF is done in an external darcs repository, and is built only as single binary.
Thus, a get-mupdf.sh script fetches it and copies a custom CMakeLists.txt for compiling a mupdf-okular library.