Use memcpy() for unaligned accesses

Use memcpy() for unaligned accesses

For unaligned memory accesses, with modern compilers memcpy() is
compiled just as efficiently as __attribute__((packed)).  This also
avoids using a nonstandard extension and potentially running into the
gcc 10 bug https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94994.