Prometheus can add latency to the data path.

Use a low priority scheduling group for the metrics.

ref: https://groups.google.com/forum/#!topic/seastar-dev/adKyzRgqc5E

> with_scheduling_group(maintenance_scheduling_group, [&] { 
>                   return 
> prometheus_server.listen(ipv4_addr{prom_addr.addr_list.front(), 
> pport}).handle_exception([pport, &cfg] (auto ep) { 
>