Refactor tabulate_rsp_biomarkers and source tabulation function h_tab_one_biomarker to avoid degenerative structure

[Melkiades]

Looking at the code, this function seems to generate a very odd structure for the table. It is based on the row_type column where it can be a split or analysis. If the latter a summarize_row_groups is done and the whole table is then do.call(rbind) together. These are both things that should be avoided. Ideally we should use splits and analysis correctly without rbind if it is from the same data-set, while, if we need to use a pre-calculated data, we should probably rely more on df_to_tt. See h_tab_one_biomarker source code for more details.