Random access papers

[bfirsh]

It should be possible to view any paper on Arxiv. This should probably be some kind of search box which converts an Arxiv URL into a web page, Scihub style. The URL /papers/[any arxiv ID] should also work.

For the backend, if a paper is accessed that doesn't exist, it should fetch metadata from arxiv then spin up a rendering job. The user could probably be shown a spinner and it could poll to refresh when it has rendered.