feat: multi-region probes via Durable Object alarms

[amondnet]

Context / Motivation

Multi-region probing is the headline differentiator competitors lead with: openstatus advertises 28-region monitoring, and Better Uptime markets multi-location checks. A single-origin probe cannot distinguish "the service is down" from "the route from one location is down", which is the exact false positive users complain about.

Cloudflare fits this well: Durable Objects with alarms give per-region probe scheduling with a location hint, so probes genuinely run from different points of presence rather than being simulated.

Proposed scope

• Durable Object-backed probe scheduling using DO alarms, with location hints to place probes across PoPs.
• Aggregate per-region latency and status into D1.
• Per-region display on the status page (latency by region, region-level status).
• Feed per-region results into quorum evaluation so a single failing region does not by itself open an incident.

Acceptance criteria

• Probes run from more than one Cloudflare location and are attributed to their region.
• Per-region latency and status are persisted and rendered.
• A single-region failure is visibly distinguished from a global outage.

Notes

Regions here map to Cloudflare points of presence, not arbitrary user-chosen geographies. Location hints are best-effort — the docs and UI must not promise a specific city.

Priority / Effort / Dependencies

• Priority: p3
• Effort: L (~1 week)
• Dependencies: none hard. Feeds the quorum logic in the alert condition DSL.