Aggregated dashboard across multiple instances

[offendingcommit]

Context

The data layer already supports an instances[] array with a single activeId (packages/web/src/lib/config.ts). With Cloud + Self-Hosted presets now in place, users can add multiple connections — but the dashboard still only ever queries the active one.

A natural follow-up is a dashboard view that fans out across all stored instances (workspaces, sessions, queue stats) and renders them side-by-side with an instance badge.

Why it's not trivial

The API client is hard-coded to the active instance (packages/web/src/api/client.ts:1-25). TanStack Query keys are entity-only (packages/web/src/api/keys.ts) — switching instances today calls qc.clear() (packages/web/src/hooks/useInstances.ts:74) precisely because the cache would otherwise collide.

To aggregate, we'd need:

1. A client factory createHonchoClient(instanceId) instead of a singleton tied to loadConfig().
2. Instance id threaded into every query key so the same workspace from two instances is cached independently.
3. New useAllWorkspaces() / useAllQueueStats() hooks that fan out across instances[] and merge results, marking each row with the originating instance.
4. Sidebar "All Instances" toggle that switches the dashboard between single-instance (current) and aggregated views.
5. UI affordance to distinguish duplicate-name entities ("workspace prod on Cloud" vs "workspace prod on Local").

Estimated 2–3 days of work.

Acceptance

• Dashboard can render workspace lists from multiple instances simultaneously
• Each row shows which instance it came from
• Queue status banner aggregates across instances when "All" is selected
• Switching back to a single instance preserves current behavior
• No cache collisions when the same workspace name exists on two instances

Related

Surfaced during review of #(PR for Honcho Cloud preset)