fix: remove duplicate AS enqueue from stale-flag reset path

When check_pending_site_created() detects a stale is_publishing flag,
it was both enqueuing a new wu_async_publish_pending_site action AND
returning publish_status=stopped to the frontend. This created two
competing retry sources: the frontend polling loop (which retries on
stopped, then reloads to kick cron) and the explicitly enqueued AS job.

Both paths can trigger publish_pending_site() concurrently, risking a
race condition where site creation is attempted twice simultaneously.

Remove the wu_enqueue_async_action call. The frontend stopped response
already handles retry (3 polls then page reload kicks cron), and the
Action Scheduler will retry on the next cron run without an explicit
enqueue.

Fixes #793

Author: superdav42

inc/managers/class-membership-manager.php

@@ -219,11 +219,13 @@ public function check_pending_site_created() {
 			);
 
 			/*
-			 * Re-enqueue the async action so Action Scheduler retries
-			 * the site creation without waiting for the next cron tick.
+			 * Return 'stopped' so the frontend retries via its own polling
+			 * loop. The Action Scheduler will also retry on the next cron
+			 * run. We do NOT enqueue an additional async action here because
+			 * that would create two competing retry sources — the frontend
+			 * poll and the AS job — which can race and attempt site creation
+			 * twice simultaneously.
 			 */
-			wu_enqueue_async_action('wu_async_publish_pending_site', ['membership_id' => $membership->get_id()], 'membership');
-
 			wp_send_json(['publish_status' => 'stopped']);
 
 			exit;