BatchedMesh: Fix toJSON, ObjectLoader integration

[gkjohnson]

Fix #30960
Fix #30961

Description

It seems BatchedMesh's toJSON and ObjectLoader integration had become out of date with some of the past changes. I tested by adding the following lines to the webgl_mesh_batch example to ensure the BatchedMesh could be cloned and serialized without error and render:

// test clone
mesh = mesh.clone();

// test serialization
new THREE.ObjectLoader().parse( mesh.toJSON(), res => {

	mesh = res;
	scene.add( mesh );

} );

A couple questions (cc @Mugen87):

• Have we considered adding toJSON and fromJSON functions to the Box3 or Sphere classes to make some of this kind of serialization more clean?
• Would we be open to adding some tests that serialize and deserialize different three.js Object3D classes and performs a diff between the source and clone to ensure the end-to-end serialization is working correctly?
• I'm sure I've asked it before but can your remind me why Object3D.toJSON handles serialization for all objects types rather than each child class implementing its own toJSON method?

[github-actions]

📦 Bundle size

Full ESM build, minified and gzipped.

	Before	After	Diff
WebGL	336.33 78.33	336.33 78.33	+0 B +0 B
WebGPU	545.96 151.35	545.96 151.35	+0 B +0 B
WebGPU Nodes	545.31 151.19	545.31 151.19	+0 B +0 B

🌳 Bundle size after tree-shaking

Minimal build including a renderer, camera, empty scene, and dependencies.

	Before	After	Diff
WebGL	465.48 112.23	465.78 112.28	+294 B +42 B
WebGPU	619.52 167.64	619.81 167.7	+294 B +55 B
WebGPU Nodes	574.39 156.92	574.68 156.98	+294 B +58 B

[Mugen87]

Have we considered adding toJSON and fromJSON functions to the Box3 or Sphere classes to make some of this kind of serialization more clean?

No that I am aware of but I that sounds like a good idea.

Normally, math objects are serialized/deserialized with toArray() and fromArray() but that is probably no good choice for more complex objects like bounding volumes.

Would we be open to adding some tests that serialize and deserialize different three.js Object3D classes and performs a diff between the source and clone to ensure the end-to-end serialization is working correctly?

The unit tests already do this partly. Example for Object3D:

three.js/test/unit/src/core/Object3D.tests.js

Line 1253 in b730a20

QUnit.test( 'toJSON', ( assert ) => {

I'm sure I've asked it before but can your remind me why Object3D.toJSON handles serialization for all objects types rather than each child class implementing its own toJSON method?

This is done for historical reasons. The project started with this approach and it has been never refactored so far. Related #11266.