Fix #19219 by moving observer triggers out of resource_scope

[mbrea-c]

Objective

Fixes #19219

Solution

Instead of calling world.commands().trigger and world.commands().trigger_targets whenever each scene is spawned, save the instance_id and optional parent entity to perform all such calls at the end. This prevents the potential flush of the world command queue that can happen if add_child is called from causing the crash.

Testing

• Did you test these changes? If so, how?
  • Verified that I can no longer reproduce the bug with the instructions at Crash when observing SceneInstanceReady trigger #19219.
  • Ran bevy_scene tests
  • Visually verified that the following examples still run as expected many_foxes, scene . (should I test any more?)
• Are there any parts that need more testing?
  • Pending to run cargo test at the root to test that all examples still build; I will update the PR when that's done
• How can other people (reviewers) test your changes? Is there anything specific they need to know?
  • Run bevy as usual
• If relevant, what platforms did you test these changes on, and are there any important ones you can't test?
  • N/a (tested on Linux/wayland but it shouldn't be relevant)

---

[mbrea-c]

No idea why CI is failing here; the errors coming up are in bevy_render and don't look related to this.

cargo run -p ci -- lints is also succeeding locally on my computer.

Am I missing something or is CI being flaky?

[greeble-dev]

The CI issue should now be resolved: #19222.