Error messaging and handling is poor for when no "static/first" codes exist in the dataset

[mmcdermott]

This should be detected early in the process and a graceful failure (or just omitting that agg as it does not actually exist) would be better.

[mmcdermott]

This is under the presumption that that agg is expected to only sometimes exist (e.g., only for static variables with a numeric value)

[mmcdermott]

See some of the error handling I added to catch this better (this is not a great example of what should be done, just an example) here: https://github.com/mmcdermott/MEDS_Tabular_AutoML/compare/55_MEDS_v03...debug_failure_on_small_mimic?expand=1