Remove extraneous fields from CSL references

[dhimmel]

Some of our methods to generate CSL items (reference metadata) produce many extraneous fields. This is most acute for Crossref DOIs, which contain many fields in addition to those part of the CSL specification. Here are some examples:

    "content-domain": {
      "domain": [],
      "crossmark-restriction": false
    },

    "link": [
      {
        "URL": "https://syndication.highwire.org/content/doi/10.1126/science.aaf5675",
        "content-type": "unspecified",
        "content-version": "vor",
        "intended-application": "similarity-checking"
      }
    ],

This creates our CSL references.json file to be unnecessarily large. Some of the fields are helpful, but are unnecessary for the purpose of creating the bibliography.

See the machine-readable schema for CSL here, which includes what fields are supported.

My proposal would be to filter (possibly optional but as default) fields using the CSL Data Schema. Potentially there would even be a way to automatically detect and delete fields that don't meet the schema to avoid downstream issues.

[dhimmel]

Seems like we'd want to use jsonschema, a python package:

• https://github.com/Julian/jsonschema
• http://python-jsonschema.readthedocs.io/en/latest/

References related to validating/correcting CSL Items based on the schema:

• http://python-jsonschema.readthedocs.io/en/latest/errors/
• https://stackoverflow.com/a/49560055/4651668
• https://python-jsonschema.readthedocs.io/en/latest/faq/#why-doesn-t-my-schema-s-default-property-set-the-default-on-my-instance