consolidated output causes publishing errors

When a user runs the analysis pipeline and the data are saved as a flat output to the results director, a `metadata.nd.json` file gets written to the same location from each instance. This will result in errors that look like this:

```
ERROR ~ Failed to publish file: s3://codeocean-s3batchbucket-16itpvq060udk/3ba99f52-878c-4133-8b98-cb19529cb391/36/c75cc66611eb7fea49b8c3bf42b252/capsule/results/metadata.nd.json; to: s3://codeocean-s3resultsbucket-1182nktl2bh9f/8ea5a002-4e1f-4b03-9260-56c1200e17f3/3ba99f52-878c-4133-8b98-cb19529cb391/2cc3b4ff-6976-43ac-a88e-040b59ba00e6/results/metadata.nd.json [copy] -- See log file for details
```

This causes the process to fail. To get around this, a user needs to partition isolate the output into separate directories