fix(synology): await producer thread shutdown before returning from uploadVideoChunks

[alexeyqu]

producerFuture.cancel(true) requests interruption but returns immediately, leaving the producer thread alive after createVideo returns. On resource-constrained CI runners, this races with Mockito's post-test stub validation, causing CreateMedia VideoModel tests to flake as SKIPPED.

Added producerFuture.get(5, TimeUnit.SECONDS) after the cancel so the method only returns once the producer has actually stopped. CancellationException and timeout are swallowed -- both mean the producer is done or on its way out.

Fixes the flaky test (see Gradle CI attempts on current master branch)

[simonxander]

The added get(5, TimeUnit.SECONDS) doesn't actually wait for the producer - it needs to come before cancel().

According to the CompletableFuture.get(long, TimeUnit) docs , get() throws CancellationException if the future was cancelled. Since producerFuture.cancel(true) runs first and completes the future into the cancelled state immediately, the subsequent get throws CancellationException right away and gets swallowed — so it returns instantly without ever waiting for the producer thread to finish. The race condition still occurs even with these changes.

Suggested fix

} finally {
  consumerAborted.set(true);
  try {
    producerFuture.get(5, TimeUnit.SECONDS);
  } catch (Exception ignored) {
    // Expected: CancellationException if cancelled, or timeout -- producer is stopping
  }
  producerFuture.cancel(true);
}

[alexeyqu]

@simonxander thank you! However, this PR didn't fix the flaky test.

After some discovery around #1495 I found the issue is test JVM OOM-ing during the test suite for Synology.

#1498 is the proposed temporary fix to unblock the CI -- and kickstart the conversation around a more permanent fix.

This PR is no longer needed, closing.