fix(configurator): expose failed staging droplet errors

[lweberru]

Findings

The failed Runner release completed image verification, lifecycle execution (exit zero), and droplet upload before CF returned a generic StagingError. Direct read-only V3 diagnostics show the linked droplet failed with StagingTimeExpired. The missing config.yml warning also occurs in the last successful release and is not proof of the failure cause.

Changes

• On Runner/API staging failure, report state, error, and timestamps of the droplet linked to the failed build rather than inspecting the current healthy droplet.
• Keep diagnostic output allowlisted, retain API router-log filtering, and explicitly report absent droplet associations. API errors and unexpected build counts fail visibly.
• Document client wait vs platform timeout and operator investigation boundaries. No automatic deployment retries, timeout changes, tag repointing or image-verification changes.

Validation

• 32/32 focused Node 24.21.0 release/buildpack/staging tests passed, including actual workflow shell executed against isolated CF fixtures for precise droplet selection, missing association, HTTP failure and private-field filtering.
• actionlint 1.7.7 passed.

This PR improves diagnosis; it does not claim to repair a server-side staging timeout. The user requested end-to-end completion of the development release through the same manual workflow available in GitHub UI. Deployment and live verification follow after CI/merge; production remains out of scope.