Runner diagnostic logs are now streamed to stdout for post-mortem debugging

Linux runners that die mid-job previously left no actionable trace once their Pod was reaped. The runner controller now sets the ACTIONS_RUNNER_PRINT_LOG_TO_STDOUT environment variable on every runner container, mirroring the actions/runner diagnostic log to stdout. That diagnostic stream flows through the existing pod-log pipeline to Loki, so the runner’s own reason for exiting survives even after the microVM and Pod are gone. This makes it possible to tell from logs whether the runner agent quit cleanly or was killed from outside the guest.