Stack monitoring errors due to missing `cluster_uuid` on `_node/pipelines?pretty&graph=true` endpoint

[robbavey]

As part of the stack monitoring with metricbeat solution, the metricbeat will poll the _node/pipelines?pretty&graph=true every 10 seconds, and send monitoring documents to elasticsearch based on the results. These results are used to generate visualizations based on the data sent from this endpoint and the _node/stats endpoint, and use the cluster_uuid to correctly correlate results and generate the appropriate visualization.

The data from _node/pipelines?pretty&graph=true is expected to be relatively static, and should only change when the pipeline itself changes. For this reason, the document id is set to be the ephemeral_id of the pipeline, and thus, the document will only change when the contents of the pipeline change.

Unfortunately, during pipeline loading, the _node/pipelines?pretty&graph=true endpoint may present an incomplete document without the cluster_uuid set, leading to issues in the stack monitoring UI, where the visualization is unable to be created, and unable to be fixed without manual intervention to delete the incomplete document.

Relates: https://github.com/elastic/kibana/issues/130217