jobsub_fetchlog should extract the logs without creating intermediate tar file
Currently, fetchlog downloads the tar/zip file from the server, extracts the contents and then deletes the tar/zip file. As per Art, the stdout/err can get really big and disk space can be an issue. Investigate options to directly stream the curl output into unzip/untar libraries so that everything happens in the memory. If such options are feasible, implementation should be hopefully straightforward.