[fix][test] Fix flaky ReplicatorTest.testReplicatorClearBacklog NPE on inFlightTask

[merlimat]

Motivation

After #25625 (Replication is stuck because failed to read entries), PersistentReplicator.readEntriesFailed now casts the ctx argument to InFlightTask and calls setEntries(Collections.emptyList()) on it. ReplicatorTest.testReplicatorClearBacklog and testReplicatorExpireMsgAsync were calling that method directly with a `null` context as a shortcut, which now NPEs:

java.lang.NullPointerException: Cannot invoke "...InFlightTask.setEntries(java.util.List)" because "inFlightTask" is null
    at o.a.p.b.s.persistent.PersistentReplicator.readEntriesFailed(PersistentReplicator.java:518)
    at o.a.p.b.s.ReplicatorTest.testReplicatorClearBacklog(ReplicatorTest.java:666)

Failing test scan: https://scans.gradle.com/s/lg3litw5ckegw/tests/task/:pulsar-broker:test/details/org.apache.pulsar.broker.service.ReplicatorTest/testReplicatorClearBacklog/2/output

Modifications

Pass a real InFlightTask instance instead of null. InFlightTask is protected in PersistentReplicator and cannot be instantiated from the org.apache.pulsar.broker.service package directly, so the construction is routed through a new helper on BrokerServicePersistInternalMethodInvoker — which is already used to expose other package-private replicator internals to tests.

Same fix is applied to testReplicatorExpireMsgAsync, which uses the identical pattern.

Verifying this change

• ReplicatorTest.testReplicatorClearBacklog and testReplicatorExpireMsgAsync pass locally with the change.

Does this pull request potentially affect one of the following parts:

• Dependencies (does it add or upgrade a dependency): no
• The public API: no
• The schema: no
• The default values of configurations: no
• The threading model: no
• The binary protocol: no
• The REST endpoints: no
• The admin CLI options: no
• The metrics: no
• Anything that affects deployment: no