fix(queue): resolve queue-worker crash-loop via Redis auth and PM2 config

Redis requires authentication but IORedis connections were not passing a
password, causing immediate NOAUTH failures and a PM2 crash-loop (1900+
restarts). Additionally, the PM2 config used `npx` as the script entry
which caused instability.

- Add REDIS_PASSWORD support to queue-service.ts and redis.ts
- Change PM2 script from npx wrapper to direct tsx CLI entry point
- Add explicit exec_mode: 'fork' to prevent cluster mode issues

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

ecosystem.config.cjs

@@ -25,8 +25,9 @@ module.exports = {
     {
       name: 'queue-worker',
       cwd: '/home/payload/payload-cms',
-      script: 'npx',
+      script: 'node_modules/tsx/dist/cli.mjs',
-      args: 'tsx scripts/run-queue-worker.ts',
+      args: 'scripts/run-queue-worker.ts',
+      exec_mode: 'fork',
       env: {
         NODE_ENV: 'production',
         // Credentials werden via dotenv aus .env geladen (siehe run-queue-worker.ts)

src/lib/queue/queue-service.ts

@@ -36,6 +36,7 @@ export function getQueueRedisConnection(): IORedis {
       host,
       port,
       db: QUEUE_REDIS_DB,
+      password: process.env.REDIS_PASSWORD || undefined,
       maxRetriesPerRequest: null, // BullMQ requirement
       enableReadyCheck: false,
     })

src/lib/redis.ts

@@ -16,6 +16,7 @@ function createRedisClient(): Redis {
     host: REDIS_HOST,
     port: REDIS_PORT,
     db: REDIS_DB,
+    password: process.env.REDIS_PASSWORD || undefined,
     maxRetriesPerRequest: 3,
     retryStrategy: (times) => {
       if (times > 3) {