Filter RemovedComponents for entities that no longer exist

[thebluefish]

Fixes #509

Here is an MVCE demonstrating a crash:

use bevy::prelude::*;
use bevy_ecs_tilemap::prelude::*;

fn main() {
    App::new()
        .add_plugins((DefaultPlugins, TilemapPlugin))
        .add_state::<ExampleState>()
        .add_systems(Startup, setup)
        .add_systems(Update, transition.run_if(in_state(ExampleState::State1)))
        .add_systems(OnExit(ExampleState::State1), despawn)
        .add_systems(OnEnter(ExampleState::State2), spawn)
        .run();
}

#[derive(Debug, Clone, Copy, Default, Eq, PartialEq, Hash, States)]
enum ExampleState {
    #[default]
    State1,
    State2,
}

fn setup(mut commands: Commands) {
    commands.spawn(Camera2dBundle::default());
    commands.spawn(TilemapBundle::default());
}

fn transition(mut next_state: ResMut<NextState<ExampleState>>) {
    println!("transitioning");
    next_state.set(ExampleState::State2);
}

fn despawn(mut commands: Commands, query: Query<Entity, With<TileStorage>>) {
    for entity in &query {
        println!("despawning {entity:?}");
        commands.entity(entity).despawn();
    }
}

fn spawn(mut commands: Commands) {
    let entity = commands.spawn_empty().id();
    println!("spawning {entity:?}");
    commands.entity(entity).insert(SpriteBundle::default());
}

transitioning
despawning 2v0
spawning 2v1
2024-02-02T18:54:12.858004Z ERROR bevy_ecs::system::commands: Failed to 'insert or spawn' bundle of type bevy_sprite::render::SpriteBatch into the following invalid entities: [2v1]
thread '' panicked at C:\Users\s33n.cargo\registry\src\index.crates.io-6f17d22bba15001f\bevy_render-0.12.1\src\render_phase\draw.rs:275:67:
called Result::unwrap() on an Err value: NoSuchEntity(2v1)

I believe the root problem is that we are erroneously inserting an Entity 2v0 into the render world via ExtractedRemovedMapBundle when that entity in our current world is now 2v1. This appears to interfere with systems that interact with that entity in the render world.

[thebluefish]

Digging into this, it seemed that my original fix was a band-aid solution that failed to understand what was being accomplished. It resolved the original problem, but simply by breaking the logic. The new solution fixes what appears to be an inconsistency when extracting, by keeping the same Entity between main and render worlds for these temp cleanup components.

[rparrett]

Have been following. This new fix makes much more sense to me too.

Confirmed all examples still function.

[rparrett]

CI is failing on fmt and clippy.

[StarArawn]

@thebluefish Can you run fmt/clippy? If not I'll branch off of your code and open a new PR. 👍 We're planning on rolling out a release tomorrow.

[thebluefish]

Ah apologies, I did not see a notification for the prior messages. Should be good now, but I'm not sure how to restart the CI.

[rparrett]

Looks good to me.