This section details on the original issue you should resolve

<issue_title>🔬 Deep-Inspection Generator Enhancement — Multi-Iteration AI Intelligence Reports</issue_title>
<issue_description>## 📋 Issue Type
Feature / Intelligence Enhancement

🎯 Objective

Enhance the deep-inspection generator (generators.ts:642-1690) to produce comprehensive AI-driven political intelligence reports with multi-iteration analysis, document cross-referencing, historical context, and predictive insights. Transform deep-inspection from a document summary tool into a full political intelligence assessment platform.

📊 Current State

• generateDeepInspection() performs single-pass document analysis
• generateDeepInspectionContent() creates 5 sections via template logic:
  1. Topic Context — static paragraph
  2. Document Intelligence — per-document 600-char analysis via generateDeepPolicyAnalysis()
  3. 5W Analysis — generateDeepAnalysisSection() cross-document synthesis
  4. Strategic Implications — template strings (only en/sv/de/fr/es)
  5. Key Takeaways — document type count bullet list
• buildDeepInspectionSections() creates 4 template sections:
  1. Stakeholder SWOT (mechanical document classification)
  2. Dashboard (single bar chart: document counts by type)
  3. Mindmap (4 static branches)
  4. Sankey (document flow diagram)
• No historical comparison or trend analysis
• No confidence scoring on analysis
• Strategic implications use hardcoded templates, not AI reasoning
• No predictive or forecasting component
• Limited to 5 languages for strategic implications (en/sv/de/fr/es)

🚀 Desired State

Deep Intelligence Report Architecture

Phase 1: Document Intelligence Collection

• Fetch documents via MCP with full-text enrichment
• AI classifies each document by: policy domain, urgency, stakeholder impact, legislative stage
• Cross-reference documents to identify related legislation, preceding motions, follow-up committee reports

Phase 2: Multi-Iteration AI Analysis

Iteration 1: Surface-level analysis (what happened)
  → Identifies key events, actors, decisions
  
Iteration 2: Deep analysis (why it matters)
  → Analyzes motivations, power dynamics, strategic implications
  
Iteration 3: Predictive analysis (what happens next)
  → Forecasts outcomes, identifies risks, suggests scenarios
  
Iteration 4: Quality review (is the analysis sound)
  → Self-reviews for bias, completeness, accuracy
  → Flags low-confidence assessments

Phase 3: Multi-Perspective Intelligence Sections

1. Executive Intelligence Summary — AI-generated 200-word overview for decision-makers
2. Document Intelligence Matrix — AI-analyzed document grid (not just titles):
   • Policy domain classification
   • Legislative stage tracking
   • Cross-reference network
   • Stakeholder impact assessment
3. Political Dynamics Analysis — AI-driven section replacing mechanical 5W:
   • Coalition stress indicators
   • Opposition strategy assessment
   • Media narrative analysis
   • Public opinion indicators
4. Predictive Assessment — New section:
   • Likely legislative outcomes (with confidence %)
   • Coalition stability forecast
   • Policy implementation timeline
   • Risk scenarios (best/worst/most-likely)
5. Historical Context — New section:
   • Precedent analysis (similar legislation outcomes)
   • Trend comparison (year-over-year changes)
   • International benchmarking (Nordic/EU comparison)
6. Enhanced Stakeholder SWOT — See 📊 Enhanced Stakeholder SWOT Analysis with AI-Generated Multi-Perspective Intelligence #1143
7. Multi-Panel Dashboard — See 📈 Interactive Dashboard Redesign with Advanced Charts and AI-Driven Metrics #1142
8. AI Conceptual Mindmap — See 🧠 Mindmap Generator Redesign with AI-Driven Conceptual Mapping and Stakeholder Perspectives #1145
9. Methodology & Confidence — New section:
   • Data sources and collection methods
   • Analysis methodology description
   • Confidence scores per section
   • Known limitations and caveats

Architecture Changes

• Refactor generateDeepInspection() into modular pipeline:

  class DeepInspectionPipeline {
    async collect(params: DeepInspectionParams): Promise<DocumentCollection>;
    async analyze(collection: DocumentCollection, iteration: number): Promise<AnalysisResult>;
    async synthesize(analyses: AnalysisResult[]): Promise<IntelligenceReport>;
    async render(report: IntelligenceReport, lang: Language): Promise<string>;
    async validate(report: IntelligenceReport): Promise<QualityScore>;
  }

• Move all deep-inspection logic from generators.ts to dedicated module:
  scripts/deep-inspection/pipeline.ts, analysis.ts, sections.ts, types.ts
• Add --depth CLI parameter: quick (1 iteration), standard (3), deep (4+)

🔧 Implementation Approach

1. Create scripts/deep-inspection/ module directory
2. Implement DeepInspectionPipeline class with configurable depth
3. Define AI prompts for each analysis iteration phase
4. Implement predictive analysis with confidence scoring
5. Add historical context via MCP document history queries
6. Create executive su...