memcached incorrectly reports NUMA mode of 'interleaved' if syscall fails

When running in an environment where the memcached process does not have sufficient privileges to set NUMA options, we don't check the request succeeded and hence print a "success" message even if it failed:

NUMA: Set memory allocation policy to 'interleave'.

By comparison, numactl prints the following:

$ docker exec 67 numactl --interleave=all bin/true

set_mempolicy: Operation not permitted

setting interleave mask: Operation not permitted

We should fix the message to make it clear the setting failed.