Refactor: Remove orphaned documentContents, improve debugging capabilities.

[adunkman]

Discovered as a part of #581.

When converting a PDF to text contents, the text contents are stored in S3. Postgres is updated with a reference to this file. There is no way to transactionalize these operations, so to avoid orphaned files in S3, we need a way to update, delete, or overwrite previously generated text contents without having associated Postgres records.

• [ ] Ensure that there are no orphaned document contents going forward

[mmarcotte]

Mike to create refactoring backlog ticket to improve debugging capabilities. Will post in here again when that's done.