enable non-data columns in aggregated boiler_fuel_eia923 table

[cmgosnell]

Right now in pudl.transform.eia923.boiler_fuel() we are dropping a lot of columns (which we want to fix because of #509 more broadly and also is necessary for the deeper fuel allocation revamp #1608 (comment). Now that we are grabbing all of the input columns on extraction (via #1836) we have the ability to keep the non-data boiler-fuel columns that are either important for analysis (re reporting_frequency_code) or useful for harvesting.

BUT this table goes through an aggregation step that right now restricts the columns to only PK or data columns in _aggregate_duplicate_boiler_fuel_keys. This was uncovered via #1836 (comment).