Consider optimizing CI builds/test runs towards changes to single benchmarks

The builds in CI are currently geared towards ensuring that changes to fuzzers/benchmarks don't break anything in production.
This is important, but it doesn't do a lot to help users that are integrating their fuzzers and or benchmarks. For example, if I integrate a new fuzzer that is broken on multiple benchmarks, I need to find/fix their build errors one by one instead of seeing all of the benchmarks its broken at once.
Ideally we could do both, so that if we detect a change in a fuzzer, we could build it with all benchmarks in parallel. This would make it much easier to debug build issues. I think this would be especially useful for new fuzzers.
The same could be done for new benchmarks. Build all fuzzers with the benchmarks at once.

At the end of the day, this may require too much maintenance (particularly considering the trouble we've had with CI so far), so we should take this into consideration.