Fuzzing fixes

[Grinkers]

Cleaned up/rebased version of #125.

This PR is mostly for the tests and history, the actual parsing implementation (clone version) will be replaced in a future PR with the EdnError/parse rework as seen in Grinkers#6. I can squash it all, but I think this will make things easier to review.

[naomijub]

Why not check the exact error?

[Grinkers]

Why not check the exact error?

For what?

Generally speaking, this is coming up next. I'm just trying to keep the PR sizes down, but the EdnError rework is pretty big
Grinkers#6 (see the "Debug" section)
https://github.com/Grinkers/edn-rs/blob/4365f555f997df08ecffce32a65c18b0dd76870d/tests/error_messages.rs
https://github.com/Grinkers/edn-rs/blob/4365f555f997df08ecffce32a65c18b0dd76870d/src/edn/error.rs

EdnError won't be string based. It also won't be possible for user/test code to be able to create an arbitrary EdnError. Nor does it have added requirements like Eq or heap allocation.

More testing can be added to the future error_messages.rs. The current tests are useful for finding regressions or parsing issues, but they're not really useful for testing inconsistent/underspecified string-based messages.

After EdnError rework, I also had some WIP instrumentation coverage testing to get 100% coverage for messages, which is where these should be tested/specified
Grinkers@eac1f8b#diff-efdbad326cffce82491ae4d979ca521416f195633dca11d3960eefaed5415a0d

[Grinkers]

Rebased and just removed the commit dealing with tests, just to get things untangled.

[Grinkers]

@naomijub I think this is a good time for a 0.17.5. Bonus points if we can do #138 at the same time to test it. This fixes a lot of things without introducing breaking changes going forward (I'm assuming the following release will probably be 0.18.x, moving towards a 1.0 goal).