Fixed an issue where `emqx_vm_cpu_use` and `emqx_vm_cpu_idle` metrics in Prometheus endpoint `/prometheus/stats` are always calculating average usage since operating system boot.