Avoid extracting joblib archives

[egibs]

We're encountering failures when trying to extract files such as:

joblib/test/data/joblib_0.9.2_compressed_pickle_py35_np19.gz

While these appear to be valid gzip archives, they are actually just application/octet-stream/data files:

joblib/test/data/joblib_0.9.2_compressed_pickle_py35_np19.gz: data

This PR avoids extracting files like this if they contain common joblib extensions and have a MIME type of application/octet-stream.

To avoid creating temporary directories (and also scanning the equivalent of "" if tmpRoot is not created), I tweaked processArchive to return early if we detect these files. The nested extraction logic has also been updated to ignore joblib archives.

[tstromberg]

I can't help but feel like hardcoding this exception introduces unnecessary complexity without a tangible benefit. Is this just to suppress output errors?

[egibs]

I can't help but feel like hardcoding this exception introduces unnecessary complexity without a tangible benefit. Is this just to suppress output errors?

Pretty much. Mainly since the files look like valid archives that should be extracting correctly.