Pre-computes sweep results for every continuous AND categorical entry filter on a block. Writes TWO sibling CSVs next to entry_filter_data.csv: entry_filter_threshold_results.csv (continuous — wide retention-target grid with tightest/max_avg variants across AvgROR/AvgPCR × low/high/combo direction) and entry_filter_categorical_results.csv (categorical — one row per category × metric, columns in_sample and out_sample for inclusion vs exclusion impact). Downstream skills (heatmap, pareto, future consumers) read these CSVs instead of recomputing. Reads only two block-local CSVs; never builds data itself. Defers upstream to alex-entry-filter-build-data.
