`bevy_ecs::world::entity_ref::EntityWorldMut::take()`, a safe function, has an undocumented unsafe block

[LikeLakers2]

Whilst writing #17335, I came across an undocumented unsafe block:

bevy/crates/bevy_ecs/src/world/entity_ref.rs

Lines 1761 to 1773 in 141b767

	#[allow(clippy::undocumented_unsafe_blocks)] // TODO: document why this is safe
	unsafe {
	Self::move_entity_from_remove::<false>(
	entity,
	&mut self.location,
	old_location.archetype_id,
	old_location,
	entities,
	archetypes,
	storages,
	new_archetype_id,
	);
	}

A safety comment should be provided. Having looked around, I found a similar call to this function, this time with a safety comment (though in an unsafe function):

bevy/crates/bevy_ecs/src/world/entity_ref.rs

Lines 1931 to 1943 in 141b767

	// SAFETY: `new_archetype_id` is a subset of the components in `old_location.archetype_id`
	// because it is created by removing a bundle from these components.
	let mut new_location = location;
	Self::move_entity_from_remove::<true>(
	entity,
	&mut new_location,
	location.archetype_id,
	location,
	&mut world.entities,
	&mut world.archetypes,
	&mut world.storages,
	new_archetype_id,
	);

which could possibly be used as a starting point for figuring out what the safety requirements of the called function even are, because the safety documentation for the called function isn't that great:

bevy/crates/bevy_ecs/src/world/entity_ref.rs

Lines 1779 to 1788 in 141b767

	/// # Safety
	///
	/// `new_archetype_id` must have the same or a subset of the components
	/// in `old_archetype_id`. Probably more safety stuff too, audit a call to
	/// this fn as if the code here was written inline
	///
	/// when DROP is true removed components will be dropped otherwise they will be forgotten
	// We use a const generic here so that we are less reliant on
	// inlining for rustc to optimize out the `match DROP`
	unsafe fn move_entity_from_remove<const DROP: bool>(

Note: While this issue is similar to #11590, I consider it different enough to make a new issue. In particular, that issue is about undocumented unsafe blocks inside unsafe functions - while this is about an undocumented unsafe block inside a safe function.

[JaySpruce]

Fixed in #18521.