**Purpose:** Add step-specific Prometheus metrics to gitlab-runner and create dashboard queries to track the script-to-steps migration progress, identify issues, and monitor operational health during rollout. **Background:** Issue #312 enables existing high-cardinality metrics (`FF_EXPORT_HIGH_CARDINALITY_METRICS=true`) for the pilot runners. This issue adds **new metric instrumentation** specific to step execution to answer: - How many jobs are using steps vs traditional execution? - Why are jobs not using steps (dispatcher decision tracking)? - Are step-specific stages (bootstrap/serve/proxy) succeeding? - What's the performance overhead of bootstrap? - Are helper binaries compatible with steps? **Metrics to Add:** **1. Job Execution Mode Tracking** `gitlab_runner_job_execution_mode_total` (Counter) - Labels: `mode={steps|traditional}`, `executor={docker|docker+machine|...}` - Purpose: Track how many jobs execute via steps vs traditional script path - Location: Increment in `common/build.go` executeScript() after dispatcher decision - **Key metric**: Adoption rate = `rate(mode="steps") / rate(total)`