* 3x encoding speedup. Finally we seem to be on par with libdv

     in terms of time and we're doing slightly better w.r.t. to
     quality. I don't think there's much room for improvement
     left, but I'd like to try and vectorize a couple of things.
     Btw, any ideas on what may impact performance will be greatly
     appreciated.

Originally committed as revision 2532 to svn://svn.ffmpeg.org/ffmpeg/trunk