Try to improve corpus pruning preprocess time.

[jonathanmetzman]

1. Limit corpus signing to 50,000 URLs.
2. Treat a lack of successful corpus pruning as a semi-failure and force removal of corpus files in this case.
3. Only fetch the names of files when only the names are used, to improve speed.

[vitorguidi]

1. Limit corpus signing to 50,000 URLs.
2. Treat a lack of successful corpus pruning as a semi-failure and force removal of corpus files in this case.
3. Only fetch the names of files when only the names are used, to improve speed.

This ultimately intends to bump up our fuzzing hours for oss fuzz, right?
Would be nice to see the profiling results that motivated this change, if you still have that 👀

[jonathanmetzman]

1. Limit corpus signing to 50,000 URLs.
2. Treat a lack of successful corpus pruning as a semi-failure and force removal of corpus files in this case.
3. Only fetch the names of files when only the names are used, to improve speed.

This ultimately intends to bump up our fuzzing hours for oss fuzz, right? Would be nice to see the profiling results that motivated this change, if you still have that 👀

This is a good point, but I don't have profiling results because the scale here was so absurd I gave up waiting. cjson_read_fuzzer had 4.5 million testcases