批量写的建议

1. PartitionGroupStoreManager.write 方法可以同时写多条消息，但为什么每处理一条消息后都会唤醒刷新进程一次，而不是把这批消息处理完成后统一唤醒一次。
2. 当前性能的瓶颈在于频繁的刷新上，而消息分组处理在ProduceManager.writeMessagesAsync, 如果可以把这部分工作后移到PartitionGroupStoreManager.write里面，那么同一分组上可以同时处理的消息会更多，它会减少flush次数，提升系统性能。

Start by tracing PartitionGroupStoreManager.write and ProduceManager.writeMessagesAsync to understand when each message wakes the flush process. Compare the current per-message behavior with batching the wake-up after all messages in a partition group are processed; done means fewer flushes while preserving batch writes and message processing.

Do mô hình lập chỉ mục viết ra từ nội dung của issue.