This also implements decoding PDF meta data strings properly, the previous way only worked by accident for non-Unicode content.
mentioned in commit 1f8aa6c9