extract author affiliations into authors.csv

[koudyk]

I just realized this doesn't work for all articles. I think different journals list the affiliations in different ways, and this is reflected in differences in the xmls

[jeromedockes]

I'm not surprised that affiliations are handled in rather diverse ways -- the archiving tag suite is meant to be very flexible so there is always a bit of variability! We'll need to look at some examples, maybe a few patterns cover the majority of articles.

Also,

• we shouldn't stuff a list of affiliations in a single csv field, instead we will need 2 files authors.csv and affiliations.csv and their rows can be linked by the author names
• the code that looks for affiliations should handle bad cases eg an author having an affiliation number that isn't anywhere in the annotation list rather than raising exceptions in those cases. pubget is sometimes used to download hundreds of thousands of papers so there are bound to be broken affiliations in the lot
• when we add tests for this they will need to check the actual values not just the fact that the field exists

I see you closed the PR; we can list the different ways we see affiliations encoded in different articles in an issue and eventually open another PR