* mln/level/transform.spe.hh: Upgrade file doc style.
	* mln/fun/v2v/linear.hh (linear_sat): New.
	
	It is very useful so that we can use both the linear effect while
	being restricted by the destination interval.
	Consider int_u12 -> int_u8.  When data are in [0,512[, the
	linear conversion function is a div 2.  Yet applying this
	function on an image can use a LUT computed from 0 to 2^12;
	unfortunately a "linear-not-saturating" function yields to
	value overflows, e.g., for 1000.
	
	* mln/level/stretch.hh (min_, max_): Fix missing update.
	(f): Use linear_sat instead of linear.
	* mln/value/lut_vec.hh (lut_vec): Fix ctor.
	(operator<<): New.


git-svn-id: https://svn.lrde.epita.fr/svn/oln/trunk@3489 4aad255d-cdde-0310-9447-f3009e2ae8c0