fix: accept StateStarting in OnStackStart lifecycle hook

Containers with Docker healthchecks show 'starting' state for
~30s after compose up. The container is connectable, just hasn't
passed its healthcheck yet. Accept both running and starting.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

controller/internal/integrations/lifecycle.go

@@ -82,14 +82,15 @@ func (m *Manager) OnStackStart(_ context.Context, stackName string) {
 			continue
 		}
 
-		// Both provider and target must be running to re-apply
+		// Both provider and target must be running (or starting) to re-apply.
+		// StateStarting = container running but healthcheck hasn't passed yet — still connectable.
 		provStack, pOk := m.stacks.GetStack(provider)
-		if !pOk || !provStack.Deployed || provStack.State != stacks.StateRunning {
+		if !pOk || !provStack.Deployed || !isStackUp(provStack.State) {
 			continue
 		}
 		if target != "filebrowser" {
 			tgtStack, tOk := m.stacks.GetStack(target)
-			if !tOk || !tgtStack.Deployed || tgtStack.State != stacks.StateRunning {
+			if !tOk || !tgtStack.Deployed || !isStackUp(tgtStack.State) {
 				continue
 			}
 		}
@@ -120,6 +121,11 @@ func (m *Manager) OnStackStart(_ context.Context, stackName string) {
 	}
 }
 
+// isStackUp returns true if the stack is running or starting (healthcheck pending).
+func isStackUp(state stacks.ContainerState) bool {
+	return state == stacks.StateRunning || state == stacks.StateStarting
+}
+
 // OnStackRemove is called when a stack is removed.
 // Permanently revokes and deletes integration state.
 func (m *Manager) OnStackRemove(_ context.Context, stackName string) {