Using Prediction Markets to Estimate the Reproducibility of Scientific Research