lavfi/mp=decimate: fix off-by-one logic in diff_C() x loop

Set x offset values in the range 0-7, rather than in the range 8-1.

The y loop is changed accordingly, to avoid confusion.

This also fixes output difference with the new pending native decimate
filter.

libavfilter/libmpcodecs/vf_decimate.c

@@ -82,8 +82,8 @@ static int diff_MMX(unsigned char *old, unsigned char *new, int os, int ns)
 static int diff_C(unsigned char *old, unsigned char *new, int os, int ns)
 {
     int x, y, d=0;
-    for (y = 8; y; y--) {
+    for (y = 0; y < 8; y++) {
-        for (x = 8; x; x--) {
+        for (x = 0; x < 8; x++) {
             d += abs(new[x] - old[x]);
         }
         new += ns;