The current acquire path calls tryAcquire() before enqueueing: TicketHolder calls tryAcquire() and only then acquire(), and UnorderedTicketSemaphore::acquire() also starts with a tryAcquire() loop. This adds extra contention and complexity.

Change the acquire path so that when a caller chooses to wait, it enqueues immediately without a prior tryAcquire(). Keep tryAcquire() as a separate non-blocking path for callers that want to avoid waiting.