Optimize data processing pipeline

Replace the use of Pandas Dataframe with Awkward arrays. 

See the [`nsbi_common_utils.datasets`](https://github.com/iris-hep/NSBI-workflow-tutorial/blob/main/src/nsbi_common_utils/datasets.py) module which loads/saves from `pandas.DataFrame`. Will also need updates to the notebook tutorials for FAIR Universe and CMS open data.