New archetypes based ECS implementation by [@zakarumych](https://github.com/zakarumych).

The novel feature of [Edict] is entity ownership implemented via reference counting.

This optional feature allows creating owned kind of `Entity` "reference",

that ensures entity is alive and despawns it on drop.

Storing owning `Entity` in the component of another entity

creates ownership relation between those entities.

Even though `Entity` is an owning reference,

components of the entity are can be queried from `World` as usual.

Shared ownership is also available.

Optimized for both high density `World`s with thousands of entities

and also for `World`s with lower number of entities spread among many archetypes.

[Edict] is aimed for wide range of game genres and use cases outside of games.

Built-in change detection with epochs allows systems to query for components

that were updated since last run of that query,

or since any other epoch as defined by `Tracks` argument.

This opens possibility to have multiple POV on changes even in single system.

For example server-side netcode can track changes individually

for each client and query for changes since last ACK.

[Edict] optimizes iteration significantly when entities

with modified components are queried.

Although ECS abbreviation implies, [Edict] does not come

with predefined `System` trait and systems scheduler.

We can call it ECQ (Entity-Component-Query) as an alternative to ECS.

[Edict] is added to [ecs_bench_suite](https://github.com/rust-gamedev/ecs_bench_suite)

so anyone can compare performance in some trivial examples with other ECS.

Development focus for February is making more public API, including unsafe parts,

to allow writing custom queries, implement schedulers with parallel execution etc.

[Edict]: https://github.com/zakarumych/edict