Fix hanging yarn test:e2e by adding proper connection lifecycle management

[Copilot]

Jest was hanging indefinitely after e2e tests because MongoDB and Redis connections were never closed — the event loop stayed alive waiting for them. --forceExit was masking the root cause.

Changes

test/app.e2e-spec.ts

• beforeEach → beforeAll; added afterAll(() => app.close()) to trigger the NestJS shutdown lifecycle

src/common/mongo/mongo.module.ts

• Splits the single factory into a MONGO_CLIENT provider (returns the MongoClient) and MONGODB derived from it
• Adds MongoClientService implements OnApplicationShutdown → client.close()
• Previously the MongoClient reference was lost inside the factory, making cleanup impossible

src/common/redis/redis.module.ts

• Adds RedisCleanupService implements OnApplicationShutdown → client.disconnect()

src/knowledgebase/websocketchat.gateway.ts

• WebSocketChatGateway creates its own private Redis client; implements OnApplicationShutdown to disconnect it

.github/workflows/package-tests.yml

• Removes --forceExit

Pattern used

@Injectable()
class MongoClientService implements OnApplicationShutdown {
  constructor(private readonly client: MongoClient) {}
  async onApplicationShutdown() {
    await this.client.close();
  }
}

NestJS calls onApplicationShutdown on all providers when app.close() is invoked, which is now guaranteed by the afterAll teardown.