Start 0.5 -> 0.6 Migration guide

[MinerSebas]

This PR starts a new Migration guide, with the following changes:

1. [Merged by Bors] - Merge AppBuilder into App bevy#2531
2. Optional .system() from [Merged by Bors] - Optional .system bevy#2398, [Merged by Bors] - Optional .system(), part 2 bevy#2403, [Merged by Bors] - Optional .system(), part 3 bevy#2422 and [Merged by Bors] - Optional .system(), part 4 (run criteria) bevy#2431,
3. [Merged by Bors] - add get_single variant bevy#2793
4. [Merged by Bors] - Rename Light => PointLight and remove unused properties bevy#1778
5. [Merged by Bors] - System Param Lifetime Split bevy#2605
6. [Merged by Bors] - Add new SystemState and rename old SystemState to SystemMeta bevy#2283
7. [Merged by Bors] - Implement and require #[derive(Component)] on all component structs bevy#2254
8. [Merged by Bors] - add documentation on Input bevy#1781

Ignored Changes:

1. [Merged by Bors] - update for wgpu 0.8 bevy#1959: Changes might be obsoleted/lumped in with the render rework
2. [Merged by Bors] - Remove with bundle filter bevy#2623: The PR has the S-Needs-Migration-Guide Label, but the feature was already broken in 0.5 and originally fixed in [Merged by Bors] - Adding WorldQuery for WithBundle bevy#2024
3. [Merged by Bors] - Change definition of ScheduleRunnerPlugin bevy#2606: The change is small and the old .add_plugin(bevy::app::ScheduleRunnerPlugin {}) syntax still compiles in 0.6.

I started this, before noticing that #134 already exists.

[alice-i-cecile]

This PR is pretty noisy because it contains a number of other commits in its history; can you rebase?

[MinerSebas]

I actually rebased the PR onto Master to get the changes from #191, as I locally use Zola 0.14.
The problem is that I am Targeting the migration-guide branch instead of master.
When the migration-guide branch was created 0.13 was still the latest version.

[jakerr]

There's a typo here it says:

Due to the System Param Lifetime Split, QuerySets now need to specify their Queries with QuerySet instead of Query .

It should say:
"... Queries with QueryState instead of Query."

[rparrett]

My rough notes from the migration process so far for a 2d game.

[ ] Handle<Texture> -> Handle<Image>
[ ] NodeBundle.material -> .color (UiColor(Color))
[ ] ImageBundle.material -> .image (Handle<Image>)
[X] SpriteBundle.material -> .texture (Handle<Image>)
[X] Visible -> Visibility 
    (and is_transparent is no longer a thing? Are things just always transparent now?)
[ ] q0_mut() -> q0(), etc
[X] SpriteBundle: Textureless sprites are spawned differently now
    (see breakout example diff from default renderer pr)
    (also cover custom_size here? when is it more appropriate to use that?)
[ ] TextureAtlasSprite: index is now usize, not u32

edit: Checking these off when PRs are opened covering them

[inodentry]

@rparrett

For 3d, is_transparent is obsoleted by the new AlphaMode field in the Standard Material, where you can choose between Opaque, Cutout (threshold value), or Blend (full transparency).

For 2d, yeah, sprites and UI are always assumed transparent. Which makes sense, because they practically always are.

[BorisBoutillier]

. mp3 is no more the default , ogg now is.
So migration of projects using mp3 must now explicitely add the features=["mp3"] in their Cargo.toml

[ramirezmike]

Things I've noticed so far that hasn't been listed

1. The casting functions for IVec, DVec, UVec, Vec have all been changed from being named after their inner elements' cast target to what the entire Vec is being casted into.

For example

// 05
let xyz: Vec3 = Vec3::new(0.0, 0.0, 0.0);
let xyz: IVec3 = xyz.as_i32();

// 06
let xyz: Vec3 = Vec3::new(0.0, 0.0, 0.0);
let xyz: IVec3 = xyz.as_ivec3();

This affects all the different dimensions of the math vectors (i.e., Vec2, Vec3 and Vec4).

2. I was going to mention that the Visible struct had been removed from TextBundle. But... actually it looks like it was re-added yesterday 👍 Just going to mention it in case someone else is not quite on latest.

3. roughness in StandardMaterial was renamed to perceptual_roughness

[rparrett]

In glam 15, PartialOrd and Ord impls were removed for vector types. Users upgrading to bevy 0.6 may see errors like this if they were trying to compare vectors.

error[E0369]: binary operation `<` cannot be applied to type `bevy::prelude::Vec2`
 --> src/main.rs:6:10
  |
6 |     if a < b {
  |        - ^ - bevy::prelude::Vec2
  |        |
  |        bevy::prelude::Vec2

In these cases, it's likely that the user was shooting themselves in the foot, because these impls were only comparing the vectors one-dimensionally.

[CGMossa]

I think the PointLight example is missing a ..Default::default() or something, as there are more fields than the ones present in the example.