Implement `Spawn` trait

[MiniaczQ]

Objective

• Implement part of #14231

Solution

• Implement Spawn trait that provides spawn_empty and spawn API
• It's implemented for World, Commands, WorldChildBuilder and ChildBuilder

Migration Guide

• Spawn was added to bevy::prelude, but will have to be included manually if prelude is not used.

[MiniaczQ]

After going through all the code for this, I think Spawn may be too specific, we could include entity(Entity) -> Self::SpawnOuyput as well, since it shares the return type, but we'd need a better name than Spawn

[MiniaczQ]

I'll fix docs after we agree on the code, it's a pain to maintain at all time

[benfrankel]

Not a full review yet but one note: Since Spawn::SpawnOutput is not given any bounds, generic contexts won't be able to do anything besides return it:

fn foo<C: Spawn>(commands: C) -> C::SpawnOutput {
    commands.spawn_empty()
        .insert(...) // Error
        .with_children(...) // Error
}

This would require another trait, so if the plan is to do that in a separate PR that makes sense.

[MiniaczQ]

I'm getting the easy traits out first, hopefully we can get a few in 0.14.1

The trait for EntityCommands and WorldMutEntity is pretty controversial imo, since the mutation would be either deferred or immediate and that's something to consider

[benfrankel]

Deferred vs immediate is already the case for Commands as Spawn vs World as Spawn, no?

[benfrankel]

Naming proposition:

• Spawn -> EntityContext
• Spawn::SpawnOutput -> EntityContext::EntityBuilder (and later this will be bound by a trait named EntityBuilder)

Or something along these lines. Issue is, EntityContext can be interpreted as either a context where you can interact with entities, OR the context of a particular entity (which would actually be EntityBuilder).

[MiniaczQ]

I want to see a more cohesive overview of which traits would be used, and what they would be implemented for. I don't think this is wise to merge piecemeal, although I am in favor of the broad idea.

If we can't do it piecewise then this will turn into a large scale PR that will need to be for 0.15, in that case I'm closing the issue in favor of future discussion in #14231 with this as an example