[fix][test] Fix flaky test ReplicatorTest.testResumptionAfterBacklogRelaxed

[void-ptr974]

Motivation

ReplicatorTest.testResumptionAfterBacklogRelaxed can time out after the remote backlog quota is relaxed. The test previously set the namespace backlog quota to 1 byte. That reliably forces the remote topic over quota, but it also leaves too little room for the pending replicated message after the existing backlog is drained. As a result, the replicator producer can keep being rejected with TopicBacklogQuotaExceededException and wait for another reconnect backoff window, causing a false test failure.

Fixes #25951.

Modifications

• Seed the remote cluster with two replicated messages and derive the test quota from the actual remote backlog size.
• Set the quota just below the current remote backlog so replication is stopped while the existing backlog is present, but can resume after the backlog is drained.
• Replace the fixed sleep with explicit Awaitility checks for the remote topic entering and leaving the backlog quota exceeded state.
• Nudge the replicator with the idempotent startProducer() call after the quota is relaxed, then wait until the replication backlog is cleared before consuming the final message.
• Keep the existing OpenTelemetry backlog and delay metric assertions around the stopped and recovered states.

Verifying this change

• ./gradlew :pulsar-broker:test --tests org.apache.pulsar.broker.service.ReplicatorTest.testResumptionAfterBacklogRelaxed
• ./gradlew :pulsar-broker:test --tests org.apache.pulsar.broker.service.ReplicatorTest.testResumptionAfterBacklogRelaxed --rerun-tasks