[ci] Add Datadog monitor definitions for unattended workflows

[eps1lon]

Adds two import-ready Datadog monitor definitions under scripts/datadog/workflow-health/ covering the workflows that run without intentional user input: the scheduled (cron) workflows and the ones driven purely by workflow_run, issues, or issue_comment events. An audit of these workflows found that several had been failing unnoticed for weeks (four of them with startup_failure since 2026-08-21), which is the gap these monitors close.

failure-monitor.json is a multi-alert ci-pipelines alert that notifies on the very first failed run of each workflow, with no consecutive-failure tolerance, and its message links the specific failed execution via the ci.pipeline.url attribute. skip-rate-monitor.json warns when a workflow's share of skipped runs over the trailing day exceeds 95 percent and alerts at 100 percent, where every completed run was skipped, which usually means a condition broke and the workflow silently stopped doing work. Both monitors keep messages minimal, notify on recovery and on no data, and renotify once a week while they stay in a triggered state. For the skip-rate monitor, whose queries match every run, the no-data path also acts as the detector for the GitHub integration itself stopping to report.

Nothing applies these definitions automatically. After merging, import each JSON via https://app.datadoghq.com/monitors/create/import and replace @slack-TODO with the real channel handle. The README lists the covered workflows (kept as a plain list so it can be re-derived from the trigger criteria) and the assumptions to verify in the CI Pipelines explorer before relying on the monitors: pipeline names match the workflow display names, skipped runs carry a distinguishable status, startup_failure runs surface as error events, and the ci.pipeline.url variable renders a run URL in the alert message.

[github-actions]

Tests Passed

Commit: 077c11b