update the averaged velocity at the beginning of the newton iteration and not at the end.

Updating at the end led to not-updating in the case of a thrown exception. If the reason for the exception was the averaged velocity itself, this value could not get rid off any more and the simulation broke down. 
reviewed by Bernd.

git-svn-id: svn://svn.iws.uni-stuttgart.de/DUMUX/dumux/trunk@12594 2fb0f335-1f38-0410-981e-8018bf24f1b0