Harden NATS messaging telemetry, scaling, and API contracts

[ORESoftware]

Scope

Audit and harden the deployed Rust NATS HTTP ingress, the always-on durable queue consumer, KEDA lag autoscaling, and the OTEL/Loki/Prometheus/API-contract integration.

The branch is being materialized by a self-validating workflow. Temporary .github/materialize/** files and the materialization workflow delete themselves only after Rust compilation, clippy, unit/live JetStream tests, deterministic OpenAPI generation, observability contract checks, and YAML validation pass.

Intended invariants

• HTTP ingress remains fail-closed and only publishes configured route/stream pairs after a JetStream ACK.
• W3C trace context crosses HTTP → NATS → consumer → downstream HTTP/NATS boundaries.
• Loki receives structured top-level service/severity/event fields.
• Prometheus directly scrapes ingress and consumer metrics and alerts on dependency loss, overload, handoff failures, fetch errors, and DLQ persistence exhaustion.
• KEDA keeps one warm listener, scales up quickly, scales down conservatively, and holds replicas when scaler telemetry fails.
• Ingress OpenAPI is generated from Rust route/schema annotations and CI rejects committed-doc drift.

Deliberately separate follow-up migrations

• Three-node JetStream R3 storage and broker HA.
• NATS TLS/accounts/subject permissions via ExternalSecrets.
• Immutable digest-pinned ingress/consumer images instead of runtime clone/build.
• Cross-replica idempotency and formal DLQ replay/redaction policy.