orca old pipeline execution cleanup vs. s3 retention policy

Issue Summary:

Orca has a mechanism for cleaning up old pipeline executions -- OldPipelineCleanupPollingNotificationAgent. One of the properties that controls that mechanism is called minimumPipelineExecutions. So, even if the general logic is to delete pipeline executions older than, say, 90 days, orca will preserve that minimum number of executions for each pipeline configuration. In other words, it's possible for very old pipeline executions to remain.

With the artifact store and its current implementation to use s3 for storage, it seems natural to configure a retention policy in s3 to delete artifacts older than a certain threshold.

Unfortunately, this means that clicking on e.g. "Baked Manifest YAML" for executions older than that threshold results in an error since s3 will have deleted the artifact from the artifact store.

cc: @xibz