Add circuit breaker for database connectivity failures during sync

[joschiv1977]

Summary

When PostgreSQL becomes unreachable during email sync (e.g. container crash, OOM kill, DNS failure), the sync loop currently continues processing every single email — logging a DB error for each one without ever stopping. This causes massive log spam, wasted Graph API / IMAP calls, and high memory/CPU usage for no benefit.

This PR adds a circuit breaker pattern to both GraphEmailService and ImapEmailService:

• Message-level circuit breaker: After 5 consecutive DB connectivity errors within a folder, abort the folder sync immediately
• Folder-level circuit breaker: After 2 consecutive folders fail due to DB issues, check database health. Abort entire account sync if DB is unreachable
• Pre-pagination health check: Before fetching the next page from Graph API, verify DB connectivity to avoid wasting API quota
• IsDbConnectivityError() helper: Distinguishes DB connectivity errors (SocketException, DNS failures, transient failures) from application-level errors (UTF-8 issues, parse errors, etc.)
• IsDatabaseReachableAsync() helper: Lightweight connectivity check via CanConnectAsync()

Non-DB errors are not affected by the circuit breaker and continue to be handled as before.

Context

This was discovered when the PostgreSQL container died during a large sync operation (800+ emails across multiple folders). The mail archiver continued running, spamming thousands of identical "Name or service not known" errors in the logs — one for every email it tried to process — while also continuing to fetch pages from the Graph API (wasting API calls). The container had to be manually stopped.

Related issues:

• [Bug]: Gmail sync loop #382 (Gmail sync loop)
• [Bug]: IMAP sync loop when emails with malformed headers fail to parse #388 (IMAP sync loop with malformed headers)
• [Bug]: could not open file "base/16384/16415": No such file or directory #363 (DB corruption after container restart)

Test plan

• Verify normal sync still works without DB issues (circuit breaker should not trigger)
• Simulate DB outage during sync (stop postgres container) — sync should abort after 5 consecutive DB errors per folder
• Verify non-DB errors (e.g. UTF-8 encoding) don't trigger circuit breaker
• Verify folder-level circuit breaker aborts account sync after 2 folder failures
• Check that sync resumes normally after DB comes back

🤖 Generated with Claude Code