refactor(jobs): add shared worker runner, locks, batching, and JSON telemetry #697

Parent epic: #679

Context

The repository has multiple one-off jobs/workers with repeated concerns: tenant iteration, retries, locking, timeouts, safe logging, dry-run behavior, and exit status.

Objective

Provide a small shared runner for scheduled/internal jobs without introducing a separate orchestration platform.

Scope

• Define job metadata, structured result, correlation/run ID, timeout, cancellation, and exit-code contract.
• Add PostgreSQL advisory-lock support and bounded tenant/item batching.
• Add dry-run support for mutating maintenance jobs where practical.
• Standardize retry classification and safe JSON telemetry.
• Keep external provider calls outside database transactions.
• Migrate two representative jobs as proof, then document incremental adoption.

Acceptance criteria

• Concurrent duplicate execution is rejected or safely skipped.
• Lock release works on success, error, timeout, and termination.
• Batches avoid unbounded transactions/memory.
• Secrets and raw personal data are redacted.
• Unit and PostgreSQL integration tests cover contention and partial failure.
• Cron/systemd/container examples remain offline/LAN compatible.