PR Description: Update Gizmo Scaling Behavior to match Ray-Based Translation

Closes #488

Problem:
The gizmo's scaling operation previously felt different in magnitude and responsiveness compared to the translation gizmo. This was because the translation gizmo had been updated to use a more accurate ray-based interaction model (as per #306), while the scaling gizmo still relied on older screen-space mouse movement calculations. This inconsistency could make precise scaling less intuitive.

Description of Changes:
This PR updates the scaling behavior of the editor gizmo to align it with the improved ray-based logic now used by the translation gizmo. The goal is to provide a more consistent and predictable user experience when manipulating objects in 3D space.

Fix Implemented:
The ApplyScale method within OvEditor::Core::GizmoBehaviour has been refactored. Instead of calculating scale based on 2D screen-space mouse displacement, it now leverages the core ray-casting and 3D plane intersection logic from the ApplyTranslation method, as suggested in the original issue (#488).

Specifically:

1. A 3D ray is cast from the current mouse position into the scene.
2. An interaction plane is defined, aligned with the gizmo's scaling axis and oriented towards the camera (similar to how the translation plane is established).
3. The intersection point of the mouse ray with this 3D plane is calculated.
4. The displacement of this intersection point on the plane (since the drag operation began) is determined.
5. This 3D displacement is then projected onto the world-space direction of the gizmo's scaling axis. The resulting scalar value represents the desired change in scale along that axis.
6. This scalar change is then applied to the object's original scale, with snapping logic and checks to prevent negative scales remaining in place.
7. The ApplyScale method is no longer const as it now modifies internal state (m_firstPick, m_initialOffset) consistent with ApplyTranslation.

By adopting this approach, the scaling operation now directly responds to how the mouse moves in the 3D world relative to the gizmo axis, rather than interpreting 2D screen movements. This should result in a scaling gizmo that feels more direct, accurate, and harmonious with the translation gizmo.

Screenshots:

before the changes in this PR:

After: