packetizer/startcode_helper: optimise AVX2 use with vzeroupper use

helps performance should SSE code also be in use at the same time

modules/packetizer/startcode_helper.h

@@ -119,6 +119,8 @@ static inline const uint8_t * startcode_FindAnnexB_AVX2( const uint8_t *p, const
             if( match & 0xF0000000 )
                 TRY_MATCH(p, 28);
         }
+
+        __asm__ volatile( "vzeroupper\n" ::: );
     }
 
     for (; p <= end; p++) {