ipatests: fix logic waiting for repl in TestIPACommand
    
    The logic of test_reset_password_unlock is twisted.
    Currently it's doing:
    - reset password on replicas[0]
    - wait for replication on master
    - kinit on master
    
    The call to wait_for_replication should be done on
    replicas[0], not on master, according to the method doc:
        Note that this waits for updates originating on this host, not those
        coming from other hosts.
    
    Fixes: https://pagure.io/freeipa/issue/8975
    
    Signed-off-by: Florence Blanc-Renaud <flo@redhat.com>
    Reviewed-By: Sergey Orlov <sorlov@redhat.com>