Emit Argo Events for failed training runs

[jmendesky]

Overview

Users of the KFP-Operator want to be informed when the training of a machine learning model has failed. This should be done similarly to how model-update events are triggered.

Notes

• change existing event name for model updates from model-update to run-completion. This will be a breaking change
• events should look as follows:

name: run-completion
payload:
  status: failed|succeeded
  pipelineName: <name of the pipeline>
  servingModelArtifact: []