[Feature] Advanced collider-level collision details

[vmarcella]

Overview

Add an advanced API that exposes collider-level contact information for users
who need per-shape detail, while keeping the default public event system
body-oriented.

Current State

The public event stream aggregates collider pairs into one RigidBody2D pair.
That is the correct default for gameplay code, but it hides which attached
shape on a compound body actually participated in contact.

Scope

Goals:

• Add an opt-in advanced collision view for per-collider detail
• Preserve the current body-oriented event stream as the default API
• Avoid leaking backend vendor types into the public surface
• Define stable handle types for colliders if needed

Non-Goals:

• Replacing the body-oriented API
• Per-substep callbacks
• Full backend manifold exposure

Proposed API

pub struct ColliderCollisionEvent {
  pub kind: CollisionEventKind,
  pub collider_a: Collider2D,
  pub collider_b: Collider2D,
  pub body_a: RigidBody2D,
  pub body_b: RigidBody2D,
  pub contact_point: Option<[f32; 2]>,
  pub normal: Option<[f32; 2]>,
  pub penetration: Option<f32>,
}

impl PhysicsWorld2D {
  pub fn collider_collision_events(
    &self,
  ) -> impl Iterator<Item = ColliderCollisionEvent>;
}

Acceptance Criteria

• A public advanced event API exists for collider-level detail
• The default body-oriented event stream remains unchanged
• Collider handles are stable and documented if newly introduced
• Compound bodies can be inspected at both body and collider granularity
• Integration tests cover multiple colliders on one body
• Docs clearly state when users should prefer body events vs.
  collider events

Affected Crates

lambda-rs, lambda-rs-platform

Notes

• This should not be treated as the primary gameplay API.
• The design is stronger if the same backend data can feed both views.