feat: add retry handling for transient sync failures

[TheZupZup]

Improves NexaNote sync reliability on unstable networks.

Adds:

• retry/backoff for transient WebDAV failures
• explicit timeout handling
• retryable sync report metadata
• safer diagnostics in sync logs
• tests for retry and non-retryable failures

This keeps the existing sync engine and file-based storage layout intact.

Details

• Retry/backoff is centralized in a single WebDAVClient._execute choke point wrapping every network op (GET, PROPFIND, PUT, MKCOL). Defaults: 3 attempts with 0.5s/1s/2s backoff. Retries only transient conditions — timeouts, connection errors, and HTTP 429/502/503/504. Auth (401/403) and 404 are never retried.
• Timeouts stay explicit on every call and are configurable via SyncConfig.timeout_seconds; the retry budget is configurable via SyncConfig.max_attempts / backoff_seconds.
• Retryable state: when a sync fails for a transient reason, the report carries retryable: true, a sanitized transient_reason, and a suggested next_retry_after_seconds. Surfaced additively on POST /sync/trigger and written to the sync log. Sync state is never corrupted.
• Diagnostics: the latest sync log now records per-operation attempt counts, the retryable flag, and the transient reason — relative DAV slug paths only, no credentials, no body content, no server URLs.

Test plan

• transient 503 succeeds after retry (PUT and GET)
• 401/404 are not retried (single attempt)
• timeout/connection error retried then reported retryable
• retry failure does not corrupt sync state (note stays LOCAL_ONLY)
• sync log records retryable state without secrets/body/URLs
• successful retry still produces a normal report
• full suite: 301 passed

Note: developed on branch claude/dreamy-brown-9HfXu per the environment's branch requirement (task suggested feat/sync-network-reliability).

---

Generated by Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3be518676b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Reset retry diagnostics before building each sync report

_collect_reliability_diagnostics() reads client.op_attempts wholesale, but that buffer is never cleared at the start of sync(). If the same NexaNoteSyncEngine instance runs multiple syncs, transient failures from an earlier run remain in op_attempts, so a later non-transient failure (for example, a 401 auth error) can be incorrectly reported as retryable=true with a stale transient reason. This can drive incorrect client retry behavior and misleading logs for long-lived engine instances.

Useful? React with 👍 / 👎.