Removed unused Neon function load_u8()

load_u8() was added in commit 7cd2428e8a80b3272046351714b26cbb88fe8d74
for use by the aom_get4x4sse_cs_neon() function, but
aom_get4x4sse_cs_neon() was removed in
commit 1d9f6897e669cef65ddcf238bfaa2960270a5001.

Bug: aomedia:3416
Change-Id: I4df386a8358e595ff61898a18e800c2a7eadac5b
diff --git a/aom_dsp/arm/mem_neon.h b/aom_dsp/arm/mem_neon.h
index 30328f7..91eb159 100644
--- a/aom_dsp/arm/mem_neon.h
+++ b/aom_dsp/arm/mem_neon.h
@@ -12,7 +12,6 @@
 #define AOM_AOM_DSP_ARM_MEM_NEON_H_
 
 #include <arm_neon.h>
-#include <assert.h>
 #include <string.h>
 #include "aom_dsp/aom_dsp_common.h"
 
@@ -101,19 +100,6 @@
   return ret;
 }
 
-// Load 2 sets of 4 bytes when alignment is guaranteed.
-static INLINE uint8x8_t load_u8(const uint8_t *buf, ptrdiff_t stride) {
-  uint32x2_t a = vdup_n_u32(0);
-
-  assert(!((intptr_t)buf % sizeof(uint32_t)));
-  assert(!(stride % sizeof(uint32_t)));
-
-  a = vld1_lane_u32((const uint32_t *)buf, a, 0);
-  buf += stride;
-  a = vld1_lane_u32((const uint32_t *)buf, a, 1);
-  return vreinterpret_u8_u32(a);
-}
-
 static INLINE void load_u8_8x8(const uint8_t *s, ptrdiff_t p,
                                uint8x8_t *const s0, uint8x8_t *const s1,
                                uint8x8_t *const s2, uint8x8_t *const s3,