feat(job-queue): resilient background job retry & monitoring (#130)

[macan88]

Problem

Async API calls (dashboard, analytics, expenses) had no retry logic or observability, meaning transient network errors or 5xx responses caused immediate user-facing failures with no recovery path.

Solution

Added JobQueue and JobMonitor classes in app/src/lib/jobQueue.ts that wrap any async function with configurable exponential backoff retry (default: 3 attempts, 200 ms base delay) and lifecycle hooks (onSuccess, onRetry, onFailure, onDead). The retry policy explicitly distinguishes transient (5xx / network) from permanent (4xx) errors — permanent errors short-circuit immediately to avoid wasting retries. Jobs that exhaust all retries are placed in a dead-letter state for operator inspection without losing error context.

Testing

• Added app/src/__tests__/jobQueue.test.ts with three focused cases:
  1. Transient failure retried up to maxAttempts, resolves on success
  2. Permanent 4xx error skips retries entirely, fires onFailure once
  3. All retries exhausted → job moves to dead-letter queue, onDead fired
• defaultRetryPolicy.isRetryable unit-tested for 4xx vs 5xx/network patterns
• Tests use jest.useFakeTimers() so backoff delays do not slow CI

Documentation

• Added docs/job-queue.md covering API, retry policy customisation, dead-letter inspection, and memory management guidance.

Closes #130