fix: Resume workflow execution (#7103)

Supercedes #7051 – if OK, I'll move the base of this PR to `develop` and we can run reviews only of this one.

**What**
- Gracefully close BullMQ Worker in Redis Event Bus and Redis Workflow Engine (by @edast, @sradevski)
- Register workflows before MedusaApp is loaded*
- Introduce `onApplicationPrepareShutdown`**
- Refactor plugin resolving for reusability purposes

*We now register workflows before modules are loaded to ensure modules can run workflows as part of bootstrapping. E.g. the Redis Workflow Engine resumes workflows when it starts, which has until this change failed, because the workflows were not registered yet.

**We introduce a new hook to prepare resources for an application shutdown. E.g. closing the BullMQ worker as a preparatory step to closing the BullMQ queue. The worker will continue to process jobs while the queue is still open to receive new jobs (without processing them). 

Co-authored-by: Stevche Radevski <4820812+sradevski@users.noreply.github.com>
Co-authored-by: Darius <618221+edast@users.noreply.github.com>
Co-authored-by: Carlos R. L. Rodrigues <37986729+carlos-r-l-rodrigues@users.noreply.github.com>

packages/workflow-engine-redis/src/utils/workflow-orchestrator-storage.ts

@@ -64,9 +64,13 @@ export class RedisDistributedTransactionStorage extends DistributedTransactionSt
     )
   }
 
+  async onApplicationPrepareShutdown() {
+    // Close worker gracefully, i.e. wait for the current jobs to finish
+    await this.worker.close()
+  }
+
   async onApplicationShutdown() {
     await this.queue.close()
-    await this.worker.close(true)
   }
 
   setWorkflowOrchestratorService(workflowOrchestratorService) {