Initial Axiom dashboards

[ncolesummers]

Summary

Create the first Axiom dashboard set for Loopworks operators using the ADR 0012 metric and trace contract.

Deliverables

• Build or document reproducible dashboards for run health, approval latency, validation failures, webhook outcomes, queue and lock pressure, and model usage/cost.
• Use ADR 0012 metric names and attributes without backend-specific naming drift.
• Store dashboard setup notes, exported definitions, or links in a repo-visible docs/config location that operators can find.
• Add any portal/catalog links needed to surface the observability entry point without duplicating Axiom data in the app.

Acceptance Criteria

• Focused tests or reviewable setup validation are defined before implementation, with a clear red or missing-state record.
• Dashboard definitions or setup steps are reproducible by another maintainer.
• Dashboard links or configuration references land in docs/config where operators can find them.
• Dashboards cover run health, approval latency, validation failures, webhook outcomes, queue/lock pressure, and model usage/cost.
• No dashboard query relies on metric names outside ADR 0012.
• UI or runtime changes include bun run build; broad observability changes finish with bun run validate.

Notes

• Keep the portal focused on links and operational state; Axiom remains the dashboard backend.
• Use screenshots as validation evidence if dashboard or portal-visible UI changes are included.
• This issue depends on telemetry emission from the foundation/instrumentation follow-ups.

Refs #21
Refs ADR 0012