Add archetype cache to queries

[cart]

Currently hecs queries iterate over all archetypes to check if that archetype matches the query. For a small number of archetypes, this is probably faster, but it will get more expensive as the number of archetypes (and/or number of components per archetype) gets larger. for a game engine (which will have many archetypes), it makes sense to have a cache. we could then invalidate the cache whenever an archetype is added/removed.

[cart]

resolved by: bd8e979

[cart]

reopening because bd8e979 doesnt actually change iteration

[DJMcNab]

I think this was fixed by #863 - at least QueryIter now has an archetypes field.

[cart]

Note quite. It has an archetypes field, but it doesn't cache matches, which means we need to check every archetype every time. My bevy_ecs core rewrite adds "stateful queries" which cache these results.

[alice-i-cecile]

Is this closed by #1525 @cart?

[cart]

Yup!