src/impls.rs: don't attempt use of neon / SIMD on big-endian aarch64.

Neon / SIMD is known to be problematical in rust, ref.
rust-lang/stdarch#1484, even
though the CPU itself supports it.

Author: he32

src/impls.rs

@@ -934,7 +934,7 @@ mod simd {
         );
         #[cfg(zerocopy_aarch64_simd)]
         simd_arch_mod!(
-            #[cfg(target_arch = "aarch64")]
+            #[cfg(all(target_arch = "aarch64", target_endian = "little"))]
             aarch64, aarch64, float32x2_t, float32x4_t, float64x1_t, float64x2_t, int8x8_t, int8x8x2_t,
             int8x8x3_t, int8x8x4_t, int8x16_t, int8x16x2_t, int8x16x3_t, int8x16x4_t, int16x4_t,
             int16x8_t, int32x2_t, int32x4_t, int64x1_t, int64x2_t, poly8x8_t, poly8x8x2_t, poly8x8x3_t,
@@ -1879,7 +1879,7 @@ mod tests {
                 vector_signed_long,
                 vector_unsigned_long
             );
-            #[cfg(all(target_arch = "aarch64", zerocopy_aarch64_simd))]
+            #[cfg(all(target_arch = "aarch64", zerocopy_aarch64_simd, target_endian = "little"))]
             #[rustfmt::skip]
             test_simd_arch_mod!(
                 aarch64, float32x2_t, float32x4_t, float64x1_t, float64x2_t, int8x8_t, int8x8x2_t,