Sliding Windows Are Not the End: Exploring Full Ranking with Long-Context Large Language Models