Improve Unique[By]/Duplicates[By] debug messages

[FHTMitchell]

Updates the Debug impls of Duplicates[By] and Unique[By] to be more descriptive. Hides the implementation detail that seen is a HashMap under the hood for Unique[By] and removes the meta redirect in Duplicates[By].

Open questions:

1. Unique doesn't actually get any benefit from having an inner: UniqueBy field since it has it's own next impl. Should we just lift those fields into the struct?
2. I renamed used to seen which feels like a more correct name, but I can revert.
3. I made the *By structs use finish_non_exhaustive which adds a .. to the representation. That's not something anything using the macro gets but useful information I feel. Might be worth adding a .. option to the macro?
4. Should the values in seen in Duplicates[By] become a 2-valued enum (not a bool) to make it clearer what's going on? Probably not worth it...

 // Before

  Unique { iter: UniqueBy { iter: Iter(["b", "c"]), used: {"a": ()} } }
  UniqueBy { iter: Iter(["b", "c"]), used: {1: ()} }
  DuplicatesBy { iter: Iter([3]), meta.used: {1: false, 2: true} }
  DuplicatesBy { iter: Iter([3]), meta.used: {0: true, 1: false} }

// After

  Unique { iter: Iter(["b", "c"]), seen: {"a"} }
  UniqueBy { iter: Iter(["b", "c"]), seen: {1}, .. }
  Duplicates   { iter: Iter([3]), seen: {1: false, 2: true}, pending: 1 }
  DuplicatesBy { iter: Iter([3]), seen: {0: true, 1: false}, pending: 1, .. }

[codecov]

Codecov Report

❌ Patch coverage is 42.00000% with 29 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.57%. Comparing base (6814180) to head (19e625d).
⚠️ Report is 211 commits behind head on master.

Files with missing lines	Patch %	Lines
src/unique_impl.rs	54.54%	15 Missing ⚠️
src/duplicates_impl.rs	17.64%	14 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1121      +/-   ##
==========================================
- Coverage   94.38%   93.57%   -0.82%     
==========================================
  Files          48       52       +4     
  Lines        6665     6708      +43     
==========================================
- Hits         6291     6277      -14     
- Misses        374      431      +57     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[phimuemue]

Hi @FHTMitchell, thanks for this. I'm gonna use some maintainers discretion here:

• Let's not start writing custom Debug implementations. They are for developers, and developers can be assumed to parse them if necessary. I only want to diverge if we have evidence that developers desperately need "better" Debug implementations.
  If you want to tackle this, please adjust impl<I> Debug for ExactlyOneError<I> accordingly so that it does what everyone else is doing. (If it bothers you: Eventually ExactlyOneError should have something nicer than Option<Either<[Item;2], Item>> e.g. something like an Array<2>::Iterator.)
• used -> seen: Fine with me, but please in a separate commit. More important is uniformity between Unique and Duplicates.
• finish_non_exhaustive: If this is really important enough, then it is not specific for Unique[By] and should be part of debug_fmt_fields!. If I am not mistaken, we prefer Derive(Debug) and use debug_fmt_fields precisely where we do not have Debug for all fields. So maybe debug_fmt_fields should have it?
• bool vs bespoke enum: I just skimmed the code and I think bool is ok.