This contains two parts:
* We explicitly need to handle text barcode contents here, due to how the
existing barcode decoder API works. This can possibly be changed later, as
that separation is far less relevant with the new system.
* Inject raw data documnent nodes for the barcode contents. This is allows
extractor scripts to recover the exact content e.g. for barcodes containing
JSON, which would otherwise be captures by the JSON-LD document processor.