Add more restrictive checks for columns of dataframes

[fhenneke]

We currently check that data frames are not missing columns we need in the accounting. For some of the transformations, it also seems important that there are not any duplicate columns.

This should be explicitly checked or handled in some other systematic way.