Sentinel test: more correct sentinels config reset.

In the initialization test for each instance we used to unregister the
old master and register it again to clear the config.
However there is a race condition doing this: as soon as we unregister
and re-register "mymaster", another Sentinel can update the new
configuration with the old state because of gossip "hello" messages.

So the correct procedure is instead, unregister "mymaster" from all the
sentinel instances, and re-register it everywhere again.

tests/sentinel/tests/includes/init-tests.tcl

@@ -12,6 +12,12 @@ test "(init) Restart killed instances" {
     }
 }
 
+test "(init) Remove old master entry from sentinels" {
+    foreach_sentinel_id id {
+        catch {S $id SENTINEL REMOVE mymaster}
+    }
+}
+
 set redis_slaves 4
 test "(init) Create a master-slaves cluster of [expr $redis_slaves+1] instances" {
     create_redis_master_slave_cluster [expr {$redis_slaves+1}]
@@ -22,7 +28,6 @@ test "(init) Sentinels can start monitoring a master" {
     set sentinels [llength $::sentinel_instances]
     set quorum [expr {$sentinels/2+1}]
     foreach_sentinel_id id {
-        catch {S $id SENTINEL REMOVE mymaster}
         S $id SENTINEL MONITOR mymaster \
               [get_instance_attrib redis $master_id host] \
               [get_instance_attrib redis $master_id port] $quorum