Runner job Metrics tab populates with CPU, memory, network, and disk usage

The Metrics tab on a runner job now shows CPU, memory, network, and disk usage sampled for the job’s full duration. Previously the collector ran on the runners-controller and could not observe usage inside the runner VMs, so every job’s Metrics tab stayed empty. Sampling has been moved into the runner itself: the macOS download fork in a shell that samples whole-VM CPU, memory, network, and Data-volume disk and POSTs every 15s; the Linux runner adds an always-on metrics sidecar that reads /proc and the JIT volume. The server endpoint now authenticates the runner Pod directly, so a token can only report metrics for its own Pod.