SRA runs should save data elsewhere and then symlink

[daler]

Currently, when running an SRA metadata sample table, the fastq files are saved directly in the data directory. However, this means that you can't delete the data directory to trigger a re-run of a workflow like you can in a non-SRA run, where it's just the symlinks living in the data directory -- because the downloaded SRA fastqs would be deleted as well.

Instead, fastq-dump should put the files in a separate directory (will need to think carefully about where this should be), and then symlink to data.