fix: use state field for ClusterShutdown instead of local variable (#458)

ClusterShutdown in ClusterManagerState was checked by update handlers to
reject work after shutdown, but was never set — only a local shouldShutdown
variable was set. This made the shutdown guard dead code, allowing updates
to slip through during the AllHandlersFinished drain. Replace shouldShutdown
with cm.state.ClusterShutdown as the single source of truth.

Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: codemonkeycxy

safe_message_handler/workflow.go

@@ -292,16 +292,15 @@ func (cm *ClusterManager) run(ctx workflow.Context) (ClusterManagerResult, error
 	cm.logger.Info("Cluster started")
 	for {
 		selector := workflow.NewSelector(ctx)
-		shouldShutdown := false
 		selector.AddReceive(cm.shutdownCh, func(c workflow.ReceiveChannel, _ bool) {
 			c.Receive(ctx, nil)
-			shouldShutdown = true
+			cm.state.ClusterShutdown = true
 		})
 		selector.AddFuture(workflow.NewTimer(ctx, cm.sleepInterval), func(f workflow.Future) {
 			cm.performHealthCheck(ctx)
 		})
 		selector.Select(ctx)
-		if shouldShutdown {
+		if cm.state.ClusterShutdown {
 			break
 		}
 		if cm.shouldContinueAsNew(ctx) {