Runner job Metrics tab populates with in-VM machine metrics

The Metrics tab on runner jobs now actually shows data. Machine metrics (CPU, memory, network, disk) are now sampled from inside the runner VM itself, for both macOS Tart runners and Linux kata runners, and reported every 15 seconds for the full duration of the job. Earlier the server-side collector produced nothing, so the Metrics tab stayed empty; users now get per-job resource usage out of the box once the new runner images are deployed.