test(canopy-eligibility): reduce orchestrator-dispatch test-process shutdown from 30s to <1s

Each orchestrator-dispatch test reports ~30s wall-clock even though the assertion fires in a few hundred milliseconds. The delay is the tokio runtime waiting for background tasks (telemetry exporters, sqlx pool drain, axum server join handles) to finish draining before the process exits.

Cosmetic-but-annoying during local dev. Find and abort (or configure shorter timeout for) whichever component holds the 30s drain window.

Extracted from orchestrator-dispatch-tests.adoc § Errata.