Advanced AI Prompt for Self-Hosted Runner Optimization

Act as a Principal Infrastructure Automation Engineer. Design a GitHub Actions workflow strictly optimized for execution on bare-metal or private cloud self-hosted runners tagged with [RUNNER_LABEL]. The workflow must: 1) Implement a robust pre-flight check evaluating disk space thresholds and zombie processes before accepting the job. 2) Execute heavy compilation tasks utilizing a strictly enforced [MAX_PARALLEL] concurrency limit to prevent runner CPU starvation. 3) Leverage persistent local NVMe filesystem caching to bypass the network latency of the standard `actions/cache`. 4) Enforce a mandatory `always()` post-build [CLEANUP_STRATEGY] that aggressively prunes dangling Docker images, drops temporary databases, and wipes the workspace to guarantee a pristine state for the next queued job.