nameserv.c: split add_domain_names() into three functions:

nameserv.c: split add_domain_names() into three functions:
            add_domain_logon_names() - adds <1c> names on WINS and broadcast.
            add_domain_master_wins() - adds <1b> name only with WINS server
            add_domain_master_bcast() - adds <1b> names only by bcast.

          : made add_domain_names() add <1c> names, unconditionally, and
            add <1b> name with the WINS server if using a WINS server, or
            add <1b> name by broadcast if _not_ using a WINS server, but
            _not_ both.

          : removed the direct parameter from remove_name_entry() and
            add_my_name_entry().

nameelect.c: made become_domain_master(), on successful registration of
             the <1b> name with the WINS server, go ahead and register
             the <1b> names by broadcast.  if the <1b> name is _not_
             successfully registered with the WINS server, the broadcast
             registration of <1b> names will _not_ proceed.

namedbsubnet.c: sorted out calls to add_my_name_entry() and remove_my_name()
                which no longer has a direct parameter.

this is all added because...

in order to fix a compatibility bug with 1.9.16p2->p11, jeremy had added a
feature that got the <1b> broadcast registered names _directly_ into the nmbd
netbios lists, undefended.  the aim was to get round the aggressive netbios
registration of <1b> names of 1.9.16p2->p11.

however, because 1.9.16p2->p11 don't properly _provide_ netlogon services,
it is better that an error message appears in 1.9.17 logs, and that
administrators are made aware of the problems with 1.9.16p2->p11, and replace
1.9.16p2->p11 servers, rather than attempt to run 1.9.16 alonside 1.9.17.

(these warning messages will need to be added...)

in _addition_, they shouldn't _be_ getting two samba servers to provide
domain logon / domain master services for the same workgroup in the same
WINS scope!

lkcl