Dashboard + observability for self-hosted instances

[JSONbored]

Part of #975.

Context

gittensory already has a dashboard. Self-hosters need visibility into what the automation is doing (activity, outcomes, health) and a place to manage configuration.

Deliverables

• The self-host stack exposes the data the dashboard needs (review/automation activity, outcomes, health) over a documented local API (auth-gated).
• Dashboard view(s) for self-hosted instances: recent activity, per-PR outcome, queue/health status.
• Basic metrics/logs surfaced (throughput, failures, provider used, fallbacks).
• (Optional) configuration management from the dashboard.

Acceptance criteria

• A self-hoster can see what the automation did and whether the instance is healthy, from the dashboard.
• No sensitive internal parameters are exposed in the UI/API.

Sub-issues

• selfhost(observability): structured per-job audit log #1202 — structured per-job audit log
• selfhost(observability): Prometheus + Grafana pre-wired observability stack #1206 — Prometheus + Grafana pre-wired observability stack
• selfhost(deploy): OpenTelemetry SDK traces + spans #1207 — OpenTelemetry SDK traces + spans
• selfhost(dashboard): dead-letter queue management UI #1208 — dead-letter queue management UI