A Bayesian model for quantifying errors in citizen science data: application to rainfall observations from Nepal