When a processor encounters an error, rather than emitting a log statement we should annotate the event with ECS error fields indicating that processing failed and why it happened. This allows querying for events that were not processed successfully. This functionality should ideally be behind a configuration flag in the shipper configuration block.

Specifically, when a processor encounters an error it should annotate the event with the event.kind: pipeline_error and add an error.message with the cause before adding it to the queue to be sent to its final destination.