This change introduce a new configuration attribute that allow to configure
the behavior of the server when replication has been configured but the
replication have not been able to connect to any of the configured replication server.

This configuration attribute is named stays in the replication domain
configuration entry and is named : ds-cfg-isolation-policy

The possible behavior values are :
- reject-all-updates : When this value is used and the server cannot connect to
 any of the replication server, all the updates are rejected with an
 UNWILLING TO PERFORM error code.
 This is the default value
- accept-all-updates :  When this value is used and the server cannot connect to
 any of the replication server, all the updates are accepted without sending them to
 a replication server, when a replication server  will become available, the server will
 use the historical information to generate replication messages.
 This last mode will make the changes at risk because they will only be saved in
 the historical information of the local server. It may also introduce high replication
 delay when the connection establish again.
 Note that currently the replication code is not yet able to generate messages for the
 MODDN, DELETE and ADD operations. (issue 1752)

This change also improve the error messages that are logged when the server cannot connect
any replication : before the server was continuously logging messages twice per seconds, now
the server only logs the error the first time and log a new message indicating that everything is
fine when a replication server is back.