Add multi-provider comparison page with side-by-side benchmarking

Steps to reproduce

When organisations need to choose between AI providers or models, they must run separate evaluations for each provider and mentally compare the results across different pages. There is no way to run the same set of scenarios against multiple providers simultaneously and see a direct visual comparison. The current dashboard shows aggregate data but does not support side-by-side benchmarking.

Proposed resolution

Add a dedicated "Compare Providers" page at `/admin/config/ai/empathy/compare`:

1. Comparison form (`AiEmpathyComparisonForm`): Multi-select for providers/models (select 2+ to compare), scenario checkboxes grouped by category, number of runs per scenario.
2. Results page (`AiEmpathyComparisonController`): Grouped bar chart comparing all 4 metrics across selected providers, radar chart overlay for each provider, data table with per-scenario breakdown.
3. Leaderboard: Rank providers by empathy performance across categories.
4. CSV/PDF export route for downloading comparison results as shareable benchmark reports.
5. Historical trends chart: Track how each provider's empathy scores change over time for model version regression detection.
6. New navigation tab "Compare" in the module's task links.

Remaining tasks

- [ ] Create comparison form with multi-provider selection
- [ ] Create comparison controller with Chart.js visualizations
- [ ] Add routing and navigation tab
- [ ] Implement batch processing for parallel provider evaluation
- [ ] Add historical trends line chart with threshold reference lines
- [ ] Add CSV export functionality
- [ ] Write tests
- [ ] Manual testing with 2+ providers