feat: implement algorithmic risk assessment engine with ML scoring and real-time monitoring

[yusuftomilola]

Summary

• Implements a TypeScript ML ensemble model (Logistic Regression + Gradient
  Boosting + Random Forest) to predict project success probability and generate
  a 0–100 risk score across six domains: Funding, Team, Technical, Community,
  Market, and Legal
• Adds SHAP-inspired explainability that attributes signed impact scores to each
  risk factor with human-readable descriptions and actionable investor recommendations
• Builds an on-chain + off-chain data pipeline with LRU caching, retry logic,
  and a privacy-preserving mode that anonymizes individual contribution data
• Introduces a real-time risk monitor using Server-Sent Events (SSE) that emits
  scored snapshots, trend analysis, and typed alerts (whale concentration,
  sentiment drops, score changes, etc.)

Changes

Core Engine (src/lib/risk-engine/)

• types.ts — Full type system: feature vectors, risk scores, alerts, monitoring
  sessions, pipeline config
• features.ts — 22-dimensional feature extraction with log-normalisation,
  sigmoid scaling, and Herfindahl concentration index
• model.ts — Pre-calibrated ensemble with 6-tree GBDT, 5-tree Random Forest,
  and logistic regression; computes 95% confidence intervals from model disagreement
• explainability.ts — SHAP marginal attribution, top risk factors, top strengths,
  explanation summary, and investor insight generation
• data-pipeline.ts — Parallel on-chain/off-chain fetching, retry-with-backoff,
  500-entry LRU cache with per-project invalidation
• monitor.ts — SSE-based polling engine with alert threshold detection, trend
  direction analysis, and pub/sub callbacks
• index.ts — RiskEngine class with assess(), assessFromRawData(), and
  assessBatch() methods

API Routes (src/app/api/risk/)

• assess/route.ts — GET + POST risk assessment with optional privacy mode
• monitor/route.ts — SSE stream for live updates + POST start/stop control
• history/[projectId]/route.ts — Paginated snapshot history

UI Components (src/components/risk/)

• RiskScoreCard — Circular overall score, success probability, confidence badge,
  and animated sub-domain progress bars

• RiskFactorChart — SHAP factor bars with current vs. benchmark comparison;
  hover reveals recommendations

• RealTimeMonitor — SSE-connected live feed with sparkline chart, trend arrow,
  and categorised alert feed

• RiskDashboard — Full composable dashboard with loading skeleton, error/retry
  state, and refresh control

  closes Algorithmic Risk Assessment Engine #76