[Test] Fix flaky test: ReaderSeekTest.testHasMessageAvailableAfterSeekToEnd

[zhanglistar]

Fixes #528

Motivation

When a reader seeks to MessageId::latest(), the broker may close the consumer and trigger a reconnect. In seekAsyncInternal, when the seek succeeds but the connection is already expired or reconnectionPending_ is true, we only set seekStatus_ = COMPLETED and did not clear the local receive queue or update startMessageId_. As a result, incomingMessages_ still held prefetched messages from before the seek until connectionOpened() later called clearReceiveQueue(). During that window, hasMessageAvailable() saw !incomingMessages_.empty() and returned true, causing the flaky test ReaderSeekTest.testHasMessageAvailableAfterSeekToEnd.

Modifications

• In ConsumerImpl::seekAsyncInternal, when seek result is OK and getCnx().expired() || reconnectionPending_ (reconnection path), we now perform the same local state updates as the non-reconnection path:
  • ackGroupingTrackerPtr_->flushAndClean()
  • incomingMessages_.clear()
  • If lastSeekArg_ holds a MessageId, set startMessageId_ from it

So after the seek is acknowledged, stale prefetched messages are cleared immediately and hasMessageAvailable() no longer returns true from old queue content before reconnect completes.

Verifying this change

• Make sure that the change passes the CI checks.

This change is already covered by existing tests:

• ReaderSeekTest.testHasMessageAvailableAfterSeekToEnd (parameterized) – asserts that after seek to latest, hasMessageAvailable() becomes false, then after producing a new message it becomes true and the message can be read; the fix removes the flakiness when the broker triggers reconnect on seek-to-end.

Documentation

• doc-not-needed

Bug fix in C++ client implementation only; no API or behavior contract change and no user-facing docs update required.