ffBase64EncodeRaw unconditionally byte-swapped the input word via
__builtin_bswap32, which only produces the intended big-endian byte
layout on little-endian hosts. On big-endian hosts the word is already
in the correct order, so the swap corrupted the encoded output.
Guard the swap with __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ so the
encoder is correct on both endiannesses.