converting the unit table to a pandas dataframe on binder takes up too much ram. I need to refit much of the code to use some other syntax. This is exacerbated by the fact that my code relies heavily on accessing the data by unit ids, which (i think) is not doable with pynwb dynamic tables.