[Feature]: Add PercentageBar annotation renderer for proportional confidence and metric indicators #103

Summary

Add a dedicated PercentageBar annotation renderer to visualize detection confidence scores, model tracking probabilities, or custom progression values as proportional metric bars anchored to detection bounding boxes.

Motivation & Use Cases

• Many computer vision applications benefit from showing model confidence or metric percentages directly on bounding boxes without cluttering text labels.
• Supports flexible placement options (Top, Bottom, InsideTop, InsideBottom), configurable dimensions, border strokes, and custom opacity.
• Preserves the session-first presentation model where renderers remain presentation-only and decoupled from core detection semantics.

Proposed Solution

1. Core Styles & Contracts:

   • Introduce PercentageBarStyle, PercentageBarPlacement, and PercentageBarDrawInstruction types.
   • Implement BasePercentageBarStyle with automatic value clamping in [0, 1], bounding box offset math, and placement alignment.
   • Provide the annotationRenderers.percentageBar({ style }) factory and presentation normalizers.

2. Vector Layer Lowering:

   • Lower PercentageBarDrawInstruction into closed background tracks and progress fill paths rendered via Pixi vector layers.
   • Wire percentageBarStyle through MediaRendererPresentation and the vector renderer layer.

3. Documentation & Playground:

   • Add a dedicated documentation page with an interactive live playground.
   • Expose playground controls (height, fill opacity, placement) in the demo workbench.