feat: Data retention policy with automated purge (UU PDP compliance)

[ajianaz]

What

Add configurable data retention policy with automated event purging.

Why

BMAD CLO finding: Error events accumulate indefinitely by default. Under UU PDP Article 16, retention must be purpose-limited and time-bound. Self-hosted users also need this to prevent unbounded storage growth.

Changes

1. Config option: TRAPFALL_RETENTION_DAYS=30 (default: 30 days, 0 = unlimited)
2. Automated purge task: Background task (tokio interval) that runs daily and deletes events older than retention period
3. Per-project override: Allow retention_days column in projects table for per-project customization
4. Purge logging: Log purge counts for observability
5. Graceful handling: Don't purge events linked to open alert rules (optional, can defer)
6. Migration: Add retention_days column to projects table (nullable, falls back to global config)

Testing

• Unit test: events older than retention → purged
• Unit test: events within retention → preserved
• Unit test: per-project override takes precedence over global config
• Integration test: purge task runs without affecting concurrent queries

[ajianaz]

Implemented in PR #305 (branch refactor/v0.3.0-ghost-crates). See commit history for details.