[mmporong]

BoundingBox.intersectsRay() gives different answers for the same ray depending on whether a hit point is requested. Without a point it uses _fastIntersectsRay() and reports a ray that starts inside the box as a hit. With a point it uses _intersectsRay(), which requires the entry distance maxMin >= 0, so the same ray returns false. OrientedBox.intersectsRay(ray, point) goes through _intersectsRay() and behaves the same way.

• _intersectsRay() now treats the box as hit unless it is entirely behind the ray (minMax >= 0), so the result no longer depends on passing point.
• For a ray that starts inside the box, point is set to where the ray leaves the box. A box entirely behind the ray still returns false (the case maxMin >= 0 was added for in Fixed infinite raycast bug in aabb intersectsRay #828).
• The JSDoc of BoundingBox.intersectsRay() and OrientedBox.intersectsRay() now describes the inside case.

Public API change (behaviour only, no signature change):

const box = new BoundingBox(new Vec3(0, 0, 0), new Vec3(1, 1, 1));
const ray = new Ray(new Vec3(0.5, 0, 0), new Vec3(1, 0, 0));
const point = new Vec3();

box.intersectsRay(ray);        // true (before and after)
box.intersectsRay(ray, point); // before: false, after: true with point = (1, 0, 0)

Tests: added #intersectsRay cases to test/core/shape/bounding-box.test.mjs (entry point from outside, exit point from inside, box behind the ray). The inside case fails before the change and passes after it, and the test/core/shape tests pass (65). I also checked OrientedBox with an identity transform: before false, after true with point (1, 0, 0). eslint is clean on the changed files. I did not run the full npm test.