Support alternate versions

[gaurav]

This PR will be used to add support for alternate versions of PubMed articles (#61). For now, it just emits URLs that include the version number (e.g. https://www.ncbi.nlm.nih.gov/pubmed/31431825.2 to indicate version 2 of PMID 31431825).

[balhoff]

Will this make it hard for someone to come to the SPARQL endpoint with a PMID, e.g. PMID:31431825 and make a query? Do they need to know to check for a version?

[gaurav]

Will this make it hard for someone to come to the SPARQL endpoint with a PMID, e.g. PMID:31431825 and make a query? Do they need to know to check for a version?

Hmm, good point. Right now, I include the original PMID with the triple: PMID:31431825.2 dcterms:isVersionOf PMID:31431825. In SPARQL, therefore, you can find all versions of a particular PMID by querying for ?article dcterms:isVersionOf PMID:31431825. I think that's probably good enough? @cbizon Will this be a problem for Robokop?

The key problem here is that AFAICT there's no way to tell whether a particular article is the most recent version or not except by looking for the largest version ID, which we can't do while processing the input as a stream. I previously considered filtering out older versions of articles when generating triples, but I'm loath to lose any potential information that might be there. I think maybe including the version number in the identifiers will make it clear that downstream users need to take the different versions into account when working with this data. For instance, I'm thinking of filtering out previous versions from the final tab-delimited output rather than messing with the triples themselves. What do you think?