feat: GNN Playground - interactive Graph Neural Network message passing visualization

Add docs/gnn-playground.html - an interactive tool for visualizing how
Graph Neural Networks propagate information through message passing:

- 7 preset graphs (Triangle, Star, Grid, Petersen, Karate Club, Cycle, K5)
- Click to add nodes, drag between nodes to add edges, right-click to delete
- Configurable aggregation (mean/sum/max), activation (ReLU/sigmoid/tanh/none)
- Adjustable layer count (1-8) and animation speed
- Animated particles showing message flow along edges
- Real-time node feature bar chart and 4x4 weight matrix display
- Layer progress indicator with step-through or auto-play modes
- Color-coded nodes reflecting current feature vectors
- Self-loop toggle for neighborhood aggregation

Usage: Open docs/gnn-playground.html in a browser, load a preset or build
a graph, configure the GNN, and step through message passing rounds.

Author: sauravbhattacharya001