Pass metadata about documents directly to the client

## Feature Request Form
### Problem you are trying to address with this feature

Presently our client application can only sense metadata for articles or documents (other than URL or PDF fingerprint) through fields in HTML pages. There is no way of detecting those metadata when our client is wrapped (for instance) around a PDF. So for instance, if someone annotates an HTML version of an article that announces a citation_doi, and then another person pulls up the PDF version, they will not see those annotations *unless* the HTML version pointed to a citation_pdf_url. And even if that were a sufficient solution, the number of publishers following absolute best practices around article metadata are low relative to the overall population.

### Your solution

Establish a way for publishers to directly pass metadata about articles and publications to our client application about the document it's viewing. In particular citation_doi and other alternate identifiers (dc:identifier, rel=alternate, etc).

This is complementary to the notion of being able to extract certain metadata elements directly from the text or metadata embedded in PDFs. A separate issue exists for that, here: https://github.com/hypothesis/product-backlog/issues/13

Start by reviewing this issue and the linked PDF metadata issue #13. Define how publishers would pass citation_doi, dc:identifier, rel=alternate, and other identifiers directly to the client, including when the client is wrapped around a PDF. Done means the interface and acceptance criteria are agreed for HTML and PDF contexts.

Written by the indexing model from the issue text.