Add artifact access via graph proxy

[yousefmoazzam]

No description provided.

[yousefmoazzam]

I'm not overly pleased about passing the argo workflows server URL from WorkflowsQuery::workflows all the way down to Task::new, it feels clumsy, but I didn't see a way to get the server URL to the Task constructor otherwise.

If there's a better way to do that then I'm happy to change this. I have a sneaking suspicion that the rejigging of the design as mentioned in the thread above may be better done sooner rather than later though, to avoid having to deal with this.

[yousefmoazzam]

If it helps checking that this works, I've been using the following query to test things:

{
  workflows(visit: {proposalCode: "mg", proposalNumber: 36964, number: 1}){
    nodes {
      name
      status {
        ...on WorkflowSucceededStatus {
          tasks {
            name,
            artifacts {
              name
              url
            }
          }
        }
      }
    }
  }
}

[garryod]

I'm not overly pleased about passing the argo workflows server URL from WorkflowsQuery::workflows all the way down to Task::new, it feels clumsy, but I didn't see a way to get the server URL to the Task constructor otherwise.

If there's a better way to do that then I'm happy to change this. I have a sneaking suspicion that the rejigging of the design as mentioned in the thread above may be better done sooner rather than later though, to avoid having to deal with this.

Another option would be to have the url field use a resolver, and store the additional information needed to construct the path in the Artifact

[yousefmoazzam]

While I'm fairly sure the filename is what we want to give requestors rather than the artifact name which argo workflows uses, maybe we'd want to provide the artifact name, so I left the two commits separate for easier viewing of using one vs. the other. Can squash if the filename is what we want.

Also, lots of assumptions being made in getting the filename:

1. argo workflow artifacts always stored in an s3 bucket (probably true, config seems to indicate this, but maybe a better way to handle this)
2. the "key" of an artifact should be non-null (seems reasonable, but I don't know much about s3 buckets, I went off of the fact that a "key" looks like a filepath relative to a bucket)
3. an artifact "key" shouldn't have .. at the end of it
4. filename should be valid utf-8

[garryod]

Looking good, could make more use of error types though, just don't want to crash in prod

[garryod]

LGTM