Implement shape-aware grip point computation for object manipulation

[Copilot]

Problem

The current implementation in reach_for_object() uses hardcoded grip point offsets like {0.1, 0.1, 0.1} and {-0.1, -0.1, 0.1} for all objects, regardless of their shape or size. This approach is suboptimal because:

• A thin book requires different grip strategies than a sphere
• Small objects may not accommodate standard grip offsets
• Large objects could benefit from multiple approach angles
• The existing vision system already provides shape information (BOX, SPHERE, CYLINDER) that wasn't being utilized

Solution

This PR implements intelligent grip point computation that adapts to object geometry:

Box Objects

Generates 3-7 grip points based on dimensions:

• Top approach (highest priority for pick-up operations)
• Side approaches when object dimensions exceed minimum grip size (4cm)
• Angled approaches for flat objects like books

Sphere Objects

Creates 5-13 grip points with circular patterns:

• Top approach with highest priority
• Equatorial ring of 8 approach points at 45° intervals
• Additional angled approaches for small spheres (<5cm radius)

Safety Constraints

• 8cm safe approach distance from object surface
• 4cm minimum grip size requirement to ensure gripper accessibility
• Gripper constraint validation before adding side approaches

Implementation Details

The solution adds three new methods to CrackleManipulation:

std::vector<Vector3> compute_grip_points_for_box(const std::vector<double>& dimensions, const geometry_msgs::msg::Pose& object_pose);
std::vector<Vector3> compute_grip_points_for_sphere(double radius, const geometry_msgs::msg::Pose& object_pose);
std::vector<Vector3> compute_shape_aware_grip_points(const moveit_msgs::msg::CollisionObject& obj);

The reach_for_object() function now calls compute_shape_aware_grip_points() instead of using hardcoded offsets, with graceful fallback to default points for unknown shapes.

Testing

Comprehensive testing shows the system generates optimal grip points:

• Standard box (10×10×5cm): 5 points (top + 4 sides)
• Thin book (20×15×2cm): 7 points (includes angled approaches)
• Small box (3×3×3cm): 3 points (respects gripper constraints)
• Ball (6cm radius): 9 points (top + equatorial pattern)
• Small ball (3cm radius): 13 points (additional angled approaches)

Backward Compatibility

✅ No breaking changes - maintains existing reach_for_object() interface
✅ Preserves all logging and ROS service functionality
✅ Graceful fallback for invalid/unknown shapes uses original default behavior
✅ Same workflow - planning and execution logic unchanged

The implementation seamlessly integrates with the existing vision system that already provides shape detection through SolidPrimitive types and dimensional information.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Compute Grip Points for objects by their shape in Crackle moveit/moveit_manipulation.cpp</issue_title>
<issue_description></issue_description>

Comments on the Issue (you are @copilot in this section)

Fixes #23

---

💬 Share your feedback on Copilot coding agent for the chance to win a $200 gift card! Click here to start the survey.