Fix neon wide loopfilter for filter8 only branch

The current code removed the check to only perform the filter8.

Change-Id: Ie54e19a77745042a5660eab986d9ef1c42e82410

vp9/common/arm/neon/vp9_mb_lpf_neon.asm

diff --git a/vp9/common/arm/neon/vp9_mb_lpf_neon.asm b/vp9/common/arm/neon/vp9_mb_lpf_neon.asm
index 8cb913c..5fe2bba 100644
--- a/vp9/common/arm/neon/vp9_mb_lpf_neon.asm
+++ b/vp9/common/arm/neon/vp9_mb_lpf_neon.asm
@@ -439,6 +439,9 @@
     tst         r7, #1
     bxne        lr
 
+    orrs        r5, r5, r6                 ; Check for 0
+    orreq       r7, r7, #2                 ; Only do mbfilter branch
+
     ; mbfilter flat && mask branch
     ; TODO(fgalligan): Can I decrease the cycles shifting to consective d's
     ; and using vibt on the q's?