Consider testing against a large sample of source code from the wild

[lread]

Originally raised by me as https://github.com/lread/rewrite-cljc-playground/issues/58

Was chatting with @sogaiu about testing against samples from the wild. He is experimenting with using clojars as a source for his projects.

Other ideas:

• cldoc as a source - selecting projects that have been successfully processed by cljdoc means we are selecting projects that have been successfully loaded. This also means we are rejecting potentially valid source that cljdoc failed to load.
• github as a source - can we search for projects that have unit tests that are currently passing? This would give some confidence that the source code is valid. Of course, it would exclude projects without unit tests.

@sogaiu pointed me to haironfire in which @jafingerhut describes how he tried to answer the question:

What fraction of Clojure projects that use Leiningen, have a project.clj file that is in EDN format, with no executable code?

and describes how he used clojars as a data source.