fix(cff): doi structure parsing

[rmfranken]

The DOI in a CFF is wrapped in an identifier object like so:

identifiers:
- type: doi
  value: 10.5281/zenodo.3555620

Our script was looking for doi objects like this:

doi: 10.5281/zenodo.3555620

Only the first case is a valid CFF format, so we must iterate through the identifier's, pick only the ones that are typed as doi, and return the value.

[rmfranken]

Only commit 4f5ef97 onwards is actually new/relevant. Soz

[cmdoret]

A few questions and comment to make sure that this is the desired behavior, but in principle it's good 👍

[rmfranken]

Thanks for your good points!

[cmdoret]

I added a commit to reduce nesting, but besides that it looks good to merge. 🙌

But please squash commits when merging this PR

[rmfranken]

Thank you for the nesting tip - looking at your change it does make sense. I was under the impression that the nesting didn't matter - if my first
try: identifiers = cff.get("identifiers", [])
would evaluate as false, that it would not continue the rest of the try block. If that's a bad assumption, I can see how splitting it into two blocks would speed it up.

also: return doi_urls or None is 🇳🇮 ce . That's neater in any case than if'ing it.