Emilie Kaufmann and her group have pointed out an error in this paper.  Specifically, the paper's Lemma 7 and 8 is likely to be false.  Unfortunately, fixing this error and obtaining the same result does not seem easy.  There might be a hope of fixing this error if there exists an algortihm whose distribution of the stopping time is exponentially decaying rather than oplynomially.  In this regard, our group is working on a paper related to this issue, so stay tuned!