Implement with_hasher adaptors

[424ever]

Closes #998

Implementations for the following methods:

• duplicates_with_hasher
• duplicates_by_with_hasher
• unique_with_hasher
• unique_by_with_hasher
• all_unique_with_hasher
• into_group_map_with_hasher
• into_group_map_by_with_hasher
• into_grouping_map_with_hasher
• into_grouping_map_by_with_hasher
• counts_with_hasher
• counts_by_with_hasher

[jswrenn]

One small nit, but otherwise this LGTM.

[phimuemue]

Two nits:

• I'm a bit unhappy about duplicated tests. We can leave them as is and add a TODO that we'd ideally want to write all test cases once and apply them to both normal and _with_hasher functions automagically.

• (Possibly over-nitpicky:) Should we rely on the fact that RandomState is the default BuildHasher or should we extract it so that std can do whatever it wants and we just adapt it? Could be done as follows:

  trait THashMap {
      type BuildHasher;
  }
  impl<K, V, S> THashMap for std::collections::HashMap<K, V, S> {
      type BuildHasher = S;
  }
  type StdDefaultBuildHasher = <std::collections::HashMap<(), ()> as THashMap>::BuildHasher;
  
  

@jswrenn Feel free to proceed as you deem appropriate.

[codecov]

Codecov Report

Attention: Patch coverage is 99.55157% with 1 line in your changes missing coverage. Please review.

Project coverage is 94.67%. Comparing base (6814180) to head (625f85f).
Report is 134 commits behind head on master.

Files with missing lines	Patch %	Lines
src/grouping_map.rs	95.45%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1007      +/-   ##
==========================================
+ Coverage   94.38%   94.67%   +0.28%     
==========================================
  Files          48       50       +2     
  Lines        6665     7062     +397     
==========================================
+ Hits         6291     6686     +395     
- Misses        374      376       +2     

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

[jswrenn]

• Should we rely on the fact that RandomState is the default BuildHasher or should we extract it so that std can do whatever it wants and we just adapt it?

It would be a breaking change for Rust to change its default hasher, so we're fine to rely on RandomState directly.

[jswrenn]

Does this buy us much? If it behaves identically to RandomState, then what's the advantage of newtyping it? I'm not worried that we've defined our with_hasher methods to literally only accept RandomState.

[424ever]

Because RandomState is the default for the type parameters, it's possible to implement the types only for S = RandomState. This is just the easiest way I thought of to catch that.

[jswrenn]

Gotcha. We can test that more directly by doing:

let _: HashMap<_, _, TestHasher> = empty::<u8>().into_group_map_with_hasher(TestHasher::new());

No need to then repeat the assertions on the output value.