feat(GraphInfluenceSeedAdvisor): agentic SPREAD/CONTAINMENT seed-selection advisor

New agentic advisor for influence maximization (SPREAD) and outbreak containment (CONTAINMENT) over undirected graphs. Uses deterministic seeded Monte Carlo of the Independent Cascade diffusion model for influence estimation, greedy marginal-gain seed selection with a degree-prefiltered candidate pool, iterative Tarjan articulation-point detection for BRIDGE_NODE/CHOKE_POINT reasoning, risk-appetite modulation (CAUTIOUS doubles sims + diversity penalty / AGGRESSIVE halves sims), SKIP_REDUNDANT alternates when marginal gain falls below threshold, A-F grade, deduped P0-first playbook (LAUNCH_SEED_CAMPAIGN / IMMUNIZE_CHOKE_POINTS / MONITOR_FRINGE / ADD_REDUNDANT_PATH with suggested (u,v) / INCREASE_BUDGET / RAISE_SIMULATIONS / BASELINE_OK), cross-seed insights (HUB_DOMINANT / DIMINISHING_RETURNS / BRIDGE_HEAVY / OUTBREAK_CONTAINED / OUTBREAK_RESILIENT), text/markdown/deterministic-JSON renderers. Pure Java 8 + JUNG, generic <V,E>, never mutates inputs. 14 passing JUnit cases.

Author: sauravbhattacharya001