Its#9468 fix hang in test079

Apparently it's possible for a new connection to be created and then the expire task to be scheduled before the first Bind operation on the new connection occurs. In that case the lc->lc_time is still 0, and conn_expire() reschedules with a negative time value, causing an infinite loop in the runq scheduler. Previously this would not happen if StartTLS was set on the connection, since ldap_prepare_conn handles that request immediately and initializes the timestamp in that case.

Now ldap_prepare_conn always inits the timestamp. Additional patch from Tero Saarni not strictly needed to fix this issue, but saves some redundant expire processing.

This supersedes !333 (closed)