Hard dependency on next-metrics and assumption of running via preflight causing problems for AWS-based software

[debugwand]

This component will fail if you don't pass in metrics, however it assumes that these metrics will be able to write to graphite at next.heroku.preflight.web_x_process_cluster_worker_y_region. This won't work for AWS.

The sticking plaster solution is to use cloudwatch in addition to this and then add each one into the healthcheck and grafana tables for monitoring.

If this needs to write to graphite it should probably be split out into an actual standalone service that is called by a client, and metrics runs on the standalone service