Aggregate code coverage of all fuzz targets per benchmark

[DonggeLiu]

Extended from #354:
Each function-under-test typically consists of 10 fuzz targets. Currently, aggregate_coverage_diff.py represents their total code coverage by the fuzz target with the highest coverage, assuming it includes all lines covered by other targets. While this assumption holds in most cases, incorporating the coverage data from all fuzz targets will yield a more accurate representation.