Better coverage benchmarks

[van Hauser]

The previous and now current coverage benchmark targets contain programs that do not show different coverage between very different fuzzers and hence do not really help the benchmark.
For SBFT23 several new programs were added which made the benchmark better.
Obviously the combined coverage benchmark was way too large, but IMHO it would be better if some of the old/current one are removed and some temporarily added from SBFT23 are permanently added.

IMHO the following targets are good for the benchmark:

assimp_assimp_fuzzer
bloaty_fuzz_target
dropbear_fuzzer-postauth_nomaths
freetype2_ftfuzzer
lcms_cms_transform_fuzzer
libpcap_fuzz_both
libxml2_xml
mbedtls_fuzz_dtlsclient
openthread_ot-ip6-send-fuzzer
proj4_proj_crs_to_crs_fuzzer
sqlite3_ossfuzz
stb_stbi_read_fuzzer
systemd_fuzz-link-parser
woff2_convert_woff2ttf_fuzzer

These are just 15 though, so 5-10 more would be good.