s3:trusts_util: make use the workstation password change more robust

s3:trusts_util: make use the workstation password change more robust

We use secrets_{prepare,failed,defer,finish}_password_change() to make
the process more robust.

Even if we just just verified the current password with the DC
it can still happen that the remote password change will fail.

If a server has the RefusePasswordChange=1 under
HKLM\SYSTEM\CurrentControlSet\Services\Netlogon\Parameters,
it will reject NetrServerPasswordSet2() with NT_STATUS_WRONG_PASSWORD.

This results in a successful local change, but a failing remote change,
which means the domain membership is broken (as we don't fallback to
the previous password for ntlmssp nor kerberos yet).

An (at least Samba) RODC will also reject a password change,
see https://bugzilla.samba.org/show_bug.cgi?id=12773.

Even with this change we still have open problems, e.g. if the password was
changed, but we didn't get the servers response. In order to fix that we need
to use only netlogon and lsa over unprotected transports, just using schannel
authentication (which supports the fallback to the old password).

BUG: https://bugzilla.samba.org/show_bug.cgi?id=12782

Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
(cherry picked from commit 40c42af11fda062fef9df96a9b5ae3e02709f07c)