Background jobs and durable file storage

## Summary
Implement durable background jobs and durable file storage suitable for self-hosted production deployments.

## Scope
- Introduce a durable job system for work that should not happen inline with user requests.
- Introduce a durable media/file storage abstraction that supports local disk and object storage.
- Move media processing, cleanup, and retryable background work onto job infrastructure.
- Make failures observable and recoverable by operators.

## Acceptance Criteria
- [ ] User-facing workflows that depend on retries or long-running work no longer require inline execution.
- [ ] Media and files survive restarts and can be served from a durable backend.
- [ ] Failed jobs can be retried and inspected by operators.
- [ ] The system has clear boundaries for what is synchronous versus background work.

## Operational Notes
- Self-hosted deployments must not depend on ephemeral local state unless explicitly configured.
- This issue should favor boring, inspectable infrastructure over clever abstractions.