10x speed up for MIME message parsing

KMime has support for a bunch of ancient legacy formats (uuencode, yenc) predating the existence of proper content type headers. Therefore the parsers for those run on every message, ie. also in the common case of regular MIME messages. And it turns out the uuencode parser is dominating the parsing cost (80+%).

See individual commits for details, after optimizing the hot path in there hit by regular MIME messages we end up with parsing those 10x faster (and needing less memory).