design is to iterate over JSON list of full entities. perform transforms/fixes.
if no changes, bail early. if changes, do a request to check that current rev
of entity is same as processed, to prevent race conditions; if a match, do
update (in import/merge batch style).

should pre-filter entities piped in. also have a CLI mode to do a single
entity; check+update code should be distinct from fix code.

releases
- extra.subtitle => subtitle
- has pmid, type is journal-article, title like "Retraction:" => type is retraction
- similar to above, title like "Retracted:" => status is retracted
- longtail release year is bogus (like > 2030?) => remove release year

files
- URL has ://archive.org/ link with rel=repository => rel=archive
- URL has ://web.archive.org/web/None/ link => delete URL
- URL has short wayback date ("2017") and another url with that as prefix => delete URL
- mimetype is bogus like (???) => clean mimetype

container
- extra.issnp = "NA" => delete key
    => in general, issne or issnp not valid ISSNs -> delete key