H264: fix signed overflow in constant multiplication

This fixes libavcodec/h264.h:1100: warning: integer overflow in expression

Originally committed as revision 22558 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/h264.h

@@ -1097,7 +1097,7 @@ static void fill_decode_caches(H264Context *h, int mb_type){
                     fill_rectangle(&h->direct_cache[scan8[0]], 4, 4, 8, MB_TYPE_16x16>>1, 1);
 
                     if(IS_DIRECT(top_type)){
-                        AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101*(MB_TYPE_DIRECT2>>1));
+                        AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101u*(MB_TYPE_DIRECT2>>1));
                     }else if(IS_8X8(top_type)){
                         int b8_xy = 4*top_xy;
                         h->direct_cache[scan8[0] + 0 - 1*8]= h->direct_table[b8_xy + 2];