[Improve] Make blocked CI triage handoffs actionable

[roomote-roomote]

​Opened on behalf of @mrubens. Follow up by mentioning @roomote-roomote, in the web UI, or in Slack.

What changed

• Require CI Failure Triage to give operators one concrete handoff when an external provider failure cannot be investigated with the task's available access.
• Name the affected system, the exact log or access needed, and the next check instead of suggesting an unverified repository change.
• Cover the external-blocker and anti-speculation requirements in the prompt contract test.

Why this change was made

CI failures caused by external deployment or infrastructure state can be correctly classified without yielding a repository fix. Those no-fix outcomes still need to leave the on-call operator with a specific next action rather than an evidence-only dead end.

Impact

CI Failure Triage remains conservative about code changes while making provider-bound incidents easier for an operator to continue. Existing repository-fix, flake, and already-green paths are unchanged.

[roomote-community]

No code issues found. See task

_{Reviewed 80af711}