This allows us not to reinterpret_cast pthread_spin_t to BAN::Atomic<pthread_spin_t> which is much better :)