pandas take pairwise difference between a lot of columns

Question

I have a pandas dataframe having a lot of actual (column name ending with _act) and projected columns (column name ending with _proj). Other than a