[summarize_sir_results] improve calculation of n_base and ref_population_pyar

[marianschmidt]

since the results created by sir_byfutime() are filtered to exclude non informative strata (i.e. no case occured in the reference population and no FU time contributed in the dataset), the summary of n_base and ref_population_pyar gives ambigous results, because for some ICD codes, single addition n will count to n_base and single strata will contribute additional ref_population_pyar. This occurs for strata where cases are observed that don't occur in refrates_df.

e.g.
Zfkd, 35-39, DE6 Hamburg, Male 2012, total_var, to 6 months, C26

This issue could be solved by

• first expanding the sir_results_df before summarizing it to get constant n_base and ref_population_pyar

or

• leaving the warning as currently implemented in summarize_sir_results()