feat: add Graph Spectral Clustering Engine - autonomous spectral partitioning

Adds GraphSpectralClusteringEngine with 7 analysis engines:
1. Laplacian Eigensolver - Jacobi rotation for normalized Laplacian eigendecomposition
2. Optimal-K Detector - eigengap heuristic with connected component detection
3. Fiedler Bisection - threshold sweep minimizing normalized cut
4. K-Way Spectral Partitioner - k-means++ on spectral embeddings with restarts
5. Conductance Analyzer - per-cluster and graph-level conductance (Cheeger approx)
6. Cluster Quality Scorer - modularity, Ncut, silhouette, composite score 0-100
7. Insight Generator - autonomous analysis of separation, balance, bridges

Features: weighted edge support, configurable k/restarts/seed, interactive
HTML dashboard with eigenvalue spectrum chart, conductance bars, quality gauge.
70 tests covering edge cases, two-clique/barbell/three-community/disconnected graphs.

Author: sauravbhattacharya001