We may need to support working with MEDS datasets that have "code modifier" columns

[mmcdermott]

These are columns that should be considered to be part of the unique codes when we tabularize and summarize the history, but in practice are stored as separate properties of each observation in the MEDS dataset. E.g., if a MEDS dataset has a code "labs//temp//F" then we can summarize that reliably in MEDS-Tab, but if a dataset has as a code "labs//temp" but then has a column "units" with value "F" vs. "C", we shouldn't ignore the units when summarizing the data.

This issue is important as many MEDS datasets will not put all relevant properties in the "code" as the examples we've been seeing do. See mmcdermott/MEDS_transforms#16 for the related MEDS transform issue.

This is pretty easy to handle just by accepting an argument of "code_modifiers" which can then be concatenated with "//"s as needed to form the codes used in MEDS-Tab