Staff-only: aggregate job failure statistics per user over a recent window.

The window is controlled by the optional ?days=N query parameter (integer). N is clamped to [1, 365]. If missing or invalid, defaults to 30 days.

Returns rows of:

• username
• total_jobs: number of jobs run
• failed_jobs: number of jobs with state OUT_OF_MEMORY or FAILED
• failed_rate: percentage of failed jobs (0-100), sorted descending. GPU fields are intentionally omitted from this initial endpoint so the page can render quickly; per-user GPU stats are fetched asynchronously from job_monitor_gpu_for_user.

Args: request (Any): Request passed to this helper.

Returns: Any: Value produced by this call (type depends on inputs).

Examples:

job_monitor(None) # doctest: +SKIP