Fix valgrind errors in the NEON loop filters.

Like the ARMv6 code, these functions were accessing values below
 the stack pointer, which can be corrupted by signal delivery at
 any time.

vp8/common/arm/neon/mbloopfilterverticaledge_uv_neon.asm

@@ -67,9 +67,10 @@
 
     sub         sp, sp, #32
     vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-    vst1.u8     {q3}, [sp]!
+    mov         r12, sp
+    vst1.u8     {q3}, [r12]!
+    vst1.u8     {q10}, [r12]!
     ldr         r12, _mbvlfuv_coeff_
-    vst1.u8     {q10}, [sp]!
 
     ; vp8_filter_mask
     vabd.u8     q11, q3, q4                 ; abs(p3 - p2)
@@ -165,8 +166,6 @@
 
     vld1.u8     {d5}, [r12]!                ;#27
 
-    sub         sp, sp, #32
-
     vqadd.s8    q6, q6, q13                 ; ps0 = vp8_signed_char_clamp(ps0 + Filter2)
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 

vp8/common/arm/neon/mbloopfilterverticaledge_y_neon.asm

@@ -63,11 +63,12 @@
     vtrn.8      q7, q8
     vtrn.8      q9, q10
 
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    vst1.u8     {q3}, [sp]!
     vld1.s8     {d4[], d5[]}, [r12]         ; thresh
+    vld1.s8     {d2[], d3[]}, [r3]          ; limit
+    mov         r12, sp
+    vst1.u8     {q3}, [r12]!
+    vst1.u8     {q10}, [r12]!
     ldr         r12, _mbvlfy_coeff_
-    vst1.u8     {q10}, [sp]!
 
     ; vp8_filter_mask
     vabd.u8     q11, q3, q4                 ; abs(p3 - p2)
@@ -150,7 +151,6 @@
 
     vld1.u8     {d6}, [r12]!                ;#18
     sub         r0, r0, r1, lsl #4
-    sub         sp, sp, #32
 
     add         r2, r0, r1