Runner job Metrics tab now shows CPU, memory, network, and disk data

Runner jobs now collect machine metrics from inside the VM instead of from the Kubernetes controller, so the Metrics tab for each job displays real CPU, memory, network, and disk usage. The previous controller-side collector could not read resource usage from Tart VMs or Kata microVMs, so the tab stayed empty. Sampling runs in the background for the full job duration and posts every 15 seconds, failing open so it never affects job execution.