Spawn entities with an EntityCommand-based abstraction

[benfrankel]

An example of how the proposed API looks in practice: https://github.com/benfrankel/bevy_jam_template/blob/main/src/screen/title.rs.

Benefits:

• You get an EntityCommands back immediately.
• You can access SystemParams ergonomically by adding them as arguments.
• You can pass in extra data (e.g. struct Projectile { lifetime, .. }).
• Converting a "pure" entity spawning system to one that takes in extra data is relatively easy:
  • Define your data type, like struct Projectile.
  • Implement EntityCommand for Projectile, and delegate to the spawning system with world.run_system_once_with((id, self), projectile).
  • Change In(id): In<Entity> to In((id, this)): In<(Entity, Projectile)>.

Drawbacks:

• Requires extension traits.
• Unfamiliar approach (using observers to spawn things is unfamiliar as well).
• spawn_fn and spawn_with unfortunately couldn't be merged into one method.
• A few lines of boilerplate to take in data. This could use a VS Code snippet to make it easier.

[janhohenheim]

I'm in favor.
Note that we went through many designs and this seems the cleanest by far imo. Bevy does not have the required answer to "How do I spawn an object composed of lots of stuff?" yet while still returning an EntityCommands, so we need to introduce an abstraction here.

[alice-i-cecile]

I'm also in favour of this.

[janhohenheim]

Blocked on upstreaming bevyengine/bevy#14278 and maybe bevyengine/bevy#14231