[Bug][Java client] Possible race condition in closing consumer which causes a leak in memory limit controller and direct memory with pooled messages

### Search before asking

- [X] I searched in the [issues](https://github.com/apache/pulsar/issues) and found nothing similar.

### Version

master

### Minimal reproduce step

This was an issue that came up in PR review comment https://github.com/apache/pulsar/pull/18928#discussion_r1049324152

### What did you expect to see?

There shouldn't be a race between org.apache.pulsar.client.impl.ConsumerImpl#closeAsync and org.apache.pulsar.client.impl.ConsumerImpl#messageReceived .

### What did you see instead?

It's possible that closeAsync and messageReceived are in a race and incoming messages don't get properly cleaned up. This can cause a leak in the memory limit controller and also leak direct memory.

### Anything else?

_No response_

### Are you willing to submit a PR?

- [ ] I'm willing to submit a PR!

Start by reading org.apache.pulsar.client.impl.ConsumerImpl#closeAsync and #messageReceived, then review the referenced PR discussion for the race scenario. The fix is complete when concurrent closing and message receipt no longer leave incoming pooled messages uncleaned or leak memory in the memory limit controller and direct memory.

Written by the indexing model from the issue text.