Hash and compare archetype identifiers by address

[Anders429]

Fixes #160.

There is currently a regression with adding and removing components via the entry API. Seems to have increased tenfold. Will have to investigate before merging.

[Anders429]

The performance regression is in the call from the entry API into Archetypes::get_mut_or_insert_new(). This call needs to change, because the identifier does not yet belong to the set of archetypes, and therefore will always create a new archetype.

[Anders429]

With the last commit, the problem is now fixed. This PR is ready to go.

[codecov-commenter]

Codecov Report

Merging #164 (79eddd8) into dev (99c3fec) will increase coverage by 0.02%.
The diff coverage is 99.76%.

@@            Coverage Diff             @@
##              dev     #164      +/-   ##
==========================================
+ Coverage   94.77%   94.80%   +0.02%     
==========================================
  Files          73       73              
  Lines       10785    10857      +72     
==========================================
+ Hits        10222    10293      +71     
- Misses        563      564       +1     

Impacted Files	Coverage Δ
src/archetypes/mod.rs	99.17% <98.83%> (-0.21%)	⬇️
src/archetype/identifier/mod.rs	95.09% <100.00%> (-0.08%)	⬇️
src/archetypes/impl_eq.rs	100.00% <100.00%> (ø)
src/archetypes/impl_serde.rs	99.39% <100.00%> (+0.01%)	⬆️
src/entity/allocator/location.rs	71.42% <100.00%> (+1.42%)	⬆️
src/entity/allocator/mod.rs	82.39% <100.00%> (+1.05%)	⬆️
src/entity/allocator/slot.rs	85.71% <100.00%> (+0.71%)	⬆️
src/world/impl_serde.rs	100.00% <100.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[Anders429]

This PR sees improvements in the old fragmented_iter benchmark from the (now archived, sadly) ecs_bench_suite. This is to be expected, since that benchmark touches a lot of archetypes, and the hashing done for each of those now costs less.

Overall, this is a good change that improves performance and removes unnecessary computations.