Refactor `CirclesArray` into a generic `GShapesArray` framework

[338rajesh]

Currently, we have a specialized CirclesArray that manages (x, y, r) with a SoA layout. While efficient, this design doesn’t scale well to other shapes (e.g., ellipses, rectangles). We should introduce a generalized GShapesArray base class to unify the handling of homogeneous shape arrays while maintaining vectorized performance.

Proposed design:

• Create GShapesArray as a generic container with three clear components:

  1. Positional parameters: (x, y) in 2D, (x, y, z) in 3D.
  2. Orientation parameters: θ in 2D, (azimuth, polar) in 3D.
  3. Size parameters: shape-specific (e.g., r for circles, (a, b) for ellipses, (w, h) for rectangles).

• Implement shared vectorized operations at the base level:

  • translate(dx, dy, dz=0)
  • rotate(angle, pivot) (2D) / spherical rotations (3D)
  • scale(factor)
  • bounding_box()

• Specialized subclasses (CirclesArray, EllipsesArray, etc.) extend this by:

  • Defining the schema of sizes
  • Implementing shape-specific operations (areas(), perimeters(), contains(), etc.)

Benefits:

• Clean separation of position/orientation vs. size.
• Consistent interface across all shapes.
• Easy extensibility for new shape families.
• Reuse of optimized vectorized transformations.

Action items:

• Draft GShapesArray base class with generic storage and transforms.
• Refactor CirclesArray to inherit from GShapesArray.
• Add EllipsesArray prototype as proof-of-concept.
• Update tests for unified API consistency.
• Benchmark performance to ensure no regressions vs. current CirclesArray.