Native messaging integration with shared job tracking

[ejsmith]

Broker-driven handlers need progress, cancellation, and history without creating a second execution scheduler. This builds on #533 by extending the normal IJobRuntimeStore, JobState, IJobMonitor, and in-memory/Redis providers. The Mediator alternative consumes these native APIs directly.

• Broker-owned jobs share monitoring, cancellation, counters, and retention with ordinary jobs. They never enter runtime claims, retry scheduling, or lease recovery. Each delivery attempt gets a fresh state-update token, and terminal state requires confirmed broker settlement.
• The messaging runtime supplies native processing context, automatic/manual delivery lease supervision, endpoint receive/drain policy, and explicit handler outcomes.
• Native AWS node subscriptions, dead-letter administration, Redis resource locks, and harness settlement observation support the sample without Mediator provider adapters.

Validation: full solution build and tests against isolated Redis and LocalStack: 2,212 passed, 24 existing capability/platform/benchmark skips. Shared provider tests cover broker/runtime isolation, stale attempts, cancellation with ambiguous acknowledgment, metadata, pagination, counters, expiry, and capacity recovery. Existing tests cover manual lease renewal, shutdown, expired history, Redis ownership, AWS fanout/cleanup, and SQS replay beyond the first receive batch.

This draft remains stacked on #533. Broker delivery and replay are at least once. AWS node resources require explicit lifecycle management; retained job history does not provide permanent deduplication or fencing of external business writes. The new APIs are unreleased and include intentional changes to the job-store contract.