Add retry logic for failed IndexedDB writes

[jwaspin]

Summary

Add best-effort retry logic for failed IndexedDB writes in v2 to reduce data loss from transient errors.

Requirements

• Wrap write ops with capped exponential backoff (e.g., ~1s, ~2s, ~4s; max 3 attempts).
• Classify errors: retry transient failures; surface non-retryable (e.g., quota exceeded) immediately with a clear UI message.
• Keep retries off the main thread; UI remains responsive.
• Ensure idempotency (no duplicate records on partial retries).
• Add lightweight logging/metrics for retry attempts and outcomes.

Acceptance Criteria

• Transient write failures are retried with capped backoff (up to 3 attempts).
• Non-retryable errors display a clear message (e.g., quota/storage full).
• No duplicate records created by the retry mechanism.
• UI stays responsive during retries; no freezes.
• Basic logs confirm retries and final status.