Maximize overall coverage by showing overlapping benchmarks

[DonggeLiu]

Extended from #354.
Context:
Currently, OSS-Fuzz-Gen selects functions with low coverage. However, some selected functions may execute the same code blocks of the project, leading to minimal project-level coverage improvement.

Goal:
Displaying the "per-project coverage increment" alongside the "per-benchmark" coverage increment will help us identify instances where overlapping code blocks are being tested. This insight will highlight cases where we should re-select benchmarks to enhance overall efficiency.

[DavidKorczynski]

Related: Fuzz Introspector has an analysis feature for "Optimal targets" which does a recursive analysis on which functions to target. Basically, pick the first best function to target, and then FI analyses the impact of this before proceeding to analyse which is the next best target to analyse. It's been around for some time and is in the UI e.g.: https://storage.googleapis.com/oss-fuzz-introspector/htslib/inspector-report/20240617/fuzz_report.html#Optimal-target-analysis

I'll expose this as an oracle API so we can use it for benchmark generation.

[DonggeLiu]

ACK, I will close this then.

[DavidKorczynski]

I still think this is a good thing to pursue though! Introspector's "Optimal targets" is just one heuristic trying to solve a similar problem, and the feature suggested in this issue is also relevant now that we can extract target functions from a diverse set of oracles.