Handling of np.nan's in the inputs for PQMass

There seems to be an odd issue if for whatever reason the inputs have np.nans, the results are skewed and non-sensical because of the np.nans. There should be a check for these nans before anything is done and correctly handling them.