Fix load from unaligned memory in aom_subtract_block_sse2()

In the function aom_subtract_block_sse2(), replace the
aligned load of pred buffer with unaligned load instruction.

BUG: aomedia:3449

Change-Id: I3a180779423c33bb1a44e3289a9fef54d9a0c37c
diff --git a/aom_dsp/x86/subtract_sse2.asm b/aom_dsp/x86/subtract_sse2.asm
index af38022..fd508c0 100644
--- a/aom_dsp/x86/subtract_sse2.asm
+++ b/aom_dsp/x86/subtract_sse2.asm
@@ -40,8 +40,8 @@
 %macro loop16 6
   mova                  m0, [srcq+%1]
   mova                  m4, [srcq+%2]
-  mova                  m1, [predq+%3]
-  mova                  m5, [predq+%4]
+  movu                  m1, [predq+%3]
+  movu                  m5, [predq+%4]
   punpckhbw             m2, m0, m7
   punpckhbw             m3, m1, m7
   punpcklbw             m0, m7