Give independently operated cron jobs independent runners

When a generic cron runner accepts one endpoint and returns one process-level result, use a separate runner instance for each independently scheduled job. This preserves per-job failure, timeout, observability, rollback, and configuration boundaries.

This applies to small systems reusing a one-shot runner. A deliberate multi-job scheduler can later be better if it provides per-job timeouts, results, and isolation. Simply looping over URLs couples failures and can delay later jobs.