use go-enry to identify generated, vendored and other types of code
I just discovered go-enry, which GitHub is using in parts of their new code search. It looks like there may be a bit there that would be useful for automatically excluding generated, vendored, and binary files. I don't think they track comment style, so we couldn't replace that, but this other stuff looks useful.
This does seem useful. Detecting vendor directories would improve walk performance (don't need to traverse the whole node_modules tree, and don't need all users to manually ignore it).
GetLanguage(filename, content) could be useful for formatting stdin to stdout (see #64) and disambiguating file types that use the same extension (.m could be Objective-C or Matlab, for example). This would require some refactoring to work with "language" as a concept rather than simple filename matching.