Correct bug in commondata loading for closure tests

Created by: andreab1997

Here we correct two bugs affecting the loading of the commondata for a closure test that uses the new names for the datasets.

The bug in loader.py is clearly a typo: new_data_stream is used instead of new_unc_stream while new_data_stream has already been closed.

The other bug is due to the fact that ds.commondata.datafile=None for new commondata. Also, in the past cd_df.iloc[cut_mask(ds.cuts), 5] was necessary to select column 5 which, as stated in the comment, is the data column. Now, using metadata.load_central(), this is no longer necessary.