[Broker] Fix direct memory leak in getLastMessageId

[lhotari]

Fixes #7304

Motivation

See #7304 for the details. There's a memory leak when calling lastMessageId API.

I used the instructions by @megfigura in #7304 (comment) to reproduce.

In one terminal window I started pulsar-perf produce to create 5 msg/second to the topic called "test"

pulsar-perf produce -r 5 -m 100000 test

Then in another window I ran the loop to call lastMessageId in a tight loop:

while true; do curl http://pulsar-testenv-deployment-broker.pulsar-testenv.svc.cluster.local:8080/admin/v2/persistent/public/default/test/lastMessageId; echo ; done

In a third terminal window, I had the broker logs filtered for errors or leak logs :

kubectl -n pulsar-testenv logs -f pod/pulsar-testenv-deployment-broker-0 |tee broker_logs_`date +%s`.txt | egrep 'ERROR|OOM|LEAK'

The test environment was deployed with these values.yaml settings: https://github.com/lhotari/pulsar-playground/blob/f39b83b9e08f80af57fb70b1439f6c466c1f4405/test-env/1node/values.yaml

The memory leak problem reproduces almost immediately.

Modifications

Add the missing entry.release() call to PersistentTopic.getLastMessageId

[eolivelli]

LGTM

@codelipenghui PTAL