Add adaptive fetcher escalation (http → browser on failure)

[gregoryfoster]

Summary

Implement automatic fetcher escalation: when the HTTP fetcher fails or produces empty/unusable content for a watch, escalate to a browser-based fetcher on the next attempt.

Context

Depends on at least one browser-based fetcher (#3 or #4) being available. The fetcher_used field on snapshots already tracks which fetcher produced each result.

Scope

• Add fetcher_preference field to Watch fetch_config (default: "auto")
• Implement escalation logic in check_watch: if HTTP fetch produces empty extraction or repeated failures, set preference to browser
• Record escalation decisions in audit log
• Allow manual override via API (user sets fetch_config.fetcher_preference = "browser")
• De-escalation: if browser fetcher succeeds with content similar to what HTTP could produce, suggest reverting

Dependencies

• Requires Add Playwright headless browser fetcher #3 (Playwright) or Add WebRecorder fetcher for archival-quality capture #4 (WebRecorder) to be implemented first

Design doc

docs/plans/2026-03-20-url-change-monitoring-design.md — Phase 7