`digmiro` is not respected when merge target is not a Sierra record

[jamieparkinson]

As documented in https://wellcome.slack.com/archives/C8X9YKM5X/p1690457327387449 and the linked thread, we encounter a bug when we merge the following:

• A Sierra record with the digmiro digcode
• MIRO images
• A CALM or TEI record for the Sierra record
• A METS record (which will be for the re-ingested MIRO images)

The intention of digmiro is to indicate that linked MIRO images have been reingested (and therefore have a METS record), meaning that the merger knows not to duplicate the corresponding images and items. However, this logic was implemented with the faulty assumption that the "target" would always be the Sierra record: this isn't true in the case that there is a CALM or a TEI record present, and we should work out how to handle that case.