Create a config to remove UNF from sample citations

Following up on discussion here:
https://groups.google.com/g/dataverse-community/c/UD3k8r8RdVE

From that discussion:

Reasons for retiring UNF

1. Standardization: UNF is only used by Dataverse repositories. As far as I know, no other repository uses it and UNF (or other hashes) are not included in recommended citations as described by Datacite or the RDA data citation working group
2. Limited use: UNF applies principally to numeric, tabular data. An increasing amount of data are not that -- this is obviously true for QDR-type qualitative data, but also applies to JSON, GIS-data, all sorts of imaging formats such as MRI data, etc.
3. False security: Since UNF only applies to tabular files, it gives a false sense of security: a dataset could, e.g., be missing an entire (non-tabular) file or have a scale description reversed in the codebook and still show up with the same UNF

Implementation details

• this should be a configuration; I'd suggest having UNF turned off by default for the reasons above, but installations may have good reasons to keep it
• As per @mercecrosas argument, the change should only apply to new datasets, as removing UNF's from existing sample citations may cause confusion. Two margin cases to consider
  1. What about datasets currently in Draft?
  2. What about datasets that are being updated

I think the cleanest solution for both of these is to keep the UNF's, so that the config simply turns UNF off for any newly initiated dataset

Start by tracing how sample citations are generated and where configuration for newly initiated datasets is defined. Done means a configuration defaults UNF off, allows installations to enable it, and preserves UNFs for existing, draft, or updated datasets as described in the issue.