improved content type guessing for RSS ("is it HTML, is it plain text? or something in between?"). It does the detection per feed now instead of per item, which is both faster and more reliable. This needed some changes behind the scenes, so item and document need private classes now. to be safe, they should be just declared for every class, regardless of whether they would be currently used in the implementation or not.

svn path=/trunk/KDE/kdepim/akregator/src/libsyndication/; revision=526501