Fuzzing improvements

[tyler92]

Major improvements:

• the fuzzer performs not only schema parsing but validation too (the most interesting part of the library)
• the fuzzer performs parsing in-memory avoiding temporary files which improves fuzzing speed
• seed corpus is based on JSON-Schema-Test-Suite/tests files

Minor improvements:

• avoid return 1 from fuzz target in case of exceptions - it makes fuzzing slower
• the line sed -i '27d' include/valijson/utils/rapidjson_utils.hpp was removed from tests/fuzzing/oss-fuzz-build.sh. A proper change is applied to the fuzzing test instead
• compile fuzzer with -Dvalijson_BUILD_TESTS=FALSE (there is no sense in compiling unit tests for fuzzing)
• added -j argument to zip to build a flat list of files (subdirectories are not consumed by the fuzzer)

List of issues found by new approach: #197 #198 #199 #200

[tristanpenman]

Thanks @tyler92, this is really interesting work! I appreciate you taking the time to raise PRs, and help improve the library.

It might take a little while for me to merge everything. I tread pretty carefully these days, as Valijson is closer to being in maintenance mode than active development. That said, I'll try to set aside some time to do so in the next week, so that we can get these fixes into a proper release.