Add round-trip parser fuzzers

[addisoncrump]

Partially addresses #14. More specialized fuzzing for e.g. correctness is necessary.

[addisoncrump]

Realized we get some round-trip tests for free here. Quickly changing these over.

[addisoncrump]

@BurntSushi This is starting to find some things, mostly inconsistencies with printing and parsing. Do you want me to investigate these and raise issues, or let you experiment with this? I need to leave this for today but can come back to it later this week (Thursday and beyond).

[BurntSushi]

If by inconsistency you mean "printing a parsed string does not return the same string back," then yeah, that's very much intended. If it's not that, then yeah, could be worth some issues. Maybe start with just one or two if there's a way to capture the general "shape" of the failures if that makes sense.

[BurntSushi]

I can also take a look as well.

[addisoncrump]

By inconsistency I mean, "We have some existing date structure and print it. Reparsing it back to date structure and comparing the date structures shows not equal."

For strtime it's a special case, I compare (parse, unparse) with (parse, unparse, parse, unparse) results since BrokenDownTime doesn't impl PartialEq.

[addisoncrump]

Maybe start with just one or two if there's a way to capture the general "shape" of the failures if that makes sense.

The example that gets spit out for temporal is: Pt843517081,1H

It is parsed, the unparsed, but then the test panics while parsing the first unparsed result because we expect that all unparsed items are parseable. The intermediate unparsed value is: PT175307616h10518456960m1774446656760s.

[BurntSushi]

By inconsistency I mean, "We have some existing date structure and print it. Reparsing it back to date structure and comparing the date structures shows not equal."

Yes, that sounds like it should be investigated!

And any panics in the parsing should definitely be considered bugs.

Probably BrokenDownTime should implement Eq, yeah.

When you get a chance, and if it's easier, filing just one issue is fine. And feel free to take your time. If there's a timeout, I'll pickup the fuzzing work from this great start. :)

[addisoncrump]

Hey, is there any interest in this still?

[BurntSushi]

Overall I'm very interested in adding fuzzing. But other work has attracted my attention.

[BurntSushi]

OK, I got this PR into a mergeable state. I mostly just cleaned the code up a bit and fixed the CI addition (your change seemed to reformat the yaml).

[BurntSushi]

Thank you!!!