cropdetect filter over-rounding down width and height.

[willo@…]

CVS version 1.10 of vf_cropdetect.c I believe mis-calculates the initial width
and height values. This was not happening to me before the "round" parameter
and code was added.

Looking at the code, I believe the lines 112 and 113 should add one onto the
difference between {x2,y2} and {x1,y1}.

As a simple example, assume that I have an NTSC disk that uses the entirity of
the screen. It would have ranges of "X: 0..719 Y:0..479". And the resulting
width and heights *should* be 720 and 480, but the code will generate values of
719 and 479 (before the rounding and trimmming steps, of course.)

I think this fix is compatable with the way the code handles all the other
rounding and shrinking.

Anyway, now I'm in the habit of looking at the raw "X: x1..x2 Y:y1..y2" numbers
and calculating it myself. But I think this should be fixed.

[reimar]

Fixed in CVS together with another error...