Add Text World Model demo scenario

Demonstrates that text-based world models need accurate bounding boxes
for collision physics. An agent navigates a room of text obstacles
(walls, signs, furniture) — Yoga's estimated text heights produce wrong
collision boundaries causing the agent to drift, while Textura's
accurate measurements produce correct physics.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: CharGrnmn