Story

As an admin I can get the following (bytes/sec) metrics for each output:

• Data sent to the forwarding destination.
• Data dropped because it could not successfully be forwarded (due to timeout, connection failure, buffers full, or any other reason)

Acceptance Criteria

• Metric data published via fluentd HTTP endpoint
• Can create a Prometheus alert using the metric (possibly combined with LOG-1032)
• Tested for accuracy against all current output types (fluentd, elastic, kafka etc...)

Notes

The requiremenst here need to be updated: see discussion on original telemetry design doc for comments on labelling and scope of these metrics, and possible re-factor of input metrics and pipeline metrics to complete the picture.