This section details on the original issue you should resolve

<issue_title>📊 Enhanced Stakeholder SWOT Analysis with AI-Generated Multi-Perspective Intelligence</issue_title>
<issue_description>## 📋 Issue Type
Feature / Architecture Enhancement

🎯 Objective

Redesign the SWOT analysis generators (swot-section.ts, stakeholder-swot-section.ts) to produce AI-driven, multi-perspective political intelligence instead of document-metadata-derived entries. Expand stakeholder perspectives from 3 to 6+, and ensure all SWOT content is generated through AI reasoning rather than mechanical document classification.

📊 Current State

• buildDeepInspectionSections() in generators.ts (lines 1102-1253) classifies documents mechanically:
  • Government strengths = propositions + laws + press releases
  • Opposition strengths = committee reports + motions
  • Private sector = hardcoded default strings
• SWOT entries are document titles truncated to 80 chars (titleOf(d).slice(0, 80))
• Impact levels (high/medium/low) assigned by document type, not actual policy impact
• Only 3 stakeholder perspectives: Government, Opposition, Private Sector
• Fallback strings in SWOT_DEFAULTS are static translations, not contextual analysis
• stakeholder-swot-section.ts renders data but doesn't analyze it

🚀 Desired State

Expanded Stakeholder Perspectives (6+)

1. Government Coalition (M, KD, L + SD support) — Policy execution, legislative agenda
2. Social Democratic Opposition (S, V, C, MP) — Scrutiny, alternative proposals
3. EU/International Actors — Regulatory alignment, diplomatic implications
4. Private Sector & Business — Regulatory impact, investment climate
5. Civil Society & NGOs — Rights impact, democratic accountability
6. Swedish Citizens/Voters — Service delivery, democratic representation

AI-Generated SWOT Content

Each SWOT entry must be:

• AI-analyzed: Generated through multi-iteration AI reasoning, not document title extraction
• Impact-assessed: AI evaluates actual policy impact (high/medium/low) with justification
• Cross-referenced: Links between related SWOT entries across stakeholders
• Trend-aware: Incorporates historical context and trend direction
• Quantified: Includes metrics where available (voting margins, budget figures, poll data)

Architecture Changes

• Create scripts/data-transformers/content-generators/ai-swot-analyzer.ts:

  interface AISwotAnalysis {
    stakeholder: string;
    perspective: StakeholderPerspective;
    strengths: AISwotEntry[];
    weaknesses: AISwotEntry[];
    opportunities: AISwotEntry[];
    threats: AISwotEntry[];
    crossReferences: SwotCrossReference[];
    confidenceScore: number; // 0-1
  }
  
  interface AISwotEntry {
    text: string;
    impact: 'high' | 'medium' | 'low';
    justification: string;
    relatedDocuments: string[];
    trendDirection: 'improving' | 'stable' | 'deteriorating';
    quantitativeEvidence?: string;
  }

• Update stakeholder-swot-section.ts to render enhanced SWOT cards with:
  • Trend indicators (↑↓→)
  • Cross-reference links between stakeholder SWOT entries
  • Confidence indicators
  • Expandable justification sections
• Remove hardcoded SWOT_DEFAULTS from generators.ts

🔧 Implementation Approach

1. Define StakeholderPerspective type with 6+ perspectives
2. Create AI analysis prompts for each stakeholder perspective
3. Implement cross-reference detection between SWOT entries
4. Add trend analysis using historical document comparison
5. Update HTML rendering with enhanced SWOT cards
6. Add CSS for trend indicators, cross-references, and confidence badges
7. Localize all 6 stakeholder names in 14 languages
8. Add unit tests for AI SWOT generation

🤖 Recommended Agent

code-quality-engineer — Refactoring SWOT generation from template-based to AI-driven

✅ Acceptance Criteria

• 6+ stakeholder perspectives (not just 3)
• All SWOT entries are AI-generated with justification
• Impact levels based on actual policy analysis, not document type
• Cross-references between related SWOT entries across stakeholders
• Trend indicators show direction (improving/stable/deteriorating)
• Confidence scores on each SWOT analysis
• No hardcoded SWOT_DEFAULTS fallback strings remain
• All 14 languages supported with localized stakeholder names
• WCAG 2.1 AA accessible SWOT cards
• CSS-only rendering (no JavaScript dependencies)
• Existing SWOT tests updated or replaced

📚 References

• Repository: https://github.com/Hack23/riksdagsmonitor
• Current SWOT: scripts/data-transformers/content-generators/swot-section.ts
• Stakeholder SWOT: scripts/data-transformers/content-generators/stakeholder-swot-section.ts
• Deep inspection SWOT builder: scripts/generate-news-enhanced/generators.ts:1102-1253
• CIA OSINT: https://github.com/Hack23/cia/blob/master/DATA_ANALYSIS_INTOP_OSINT.md

🏷️ Labels

`typ...