ENH: Speed up S3DataGrabber using prefix arg

[lukassnoek]

Changes proposed in this pull request

• When finding files on S3 (using S3DataGrabber), the command ...

  bkt_files = list(k.key for k in bkt.list())

... finds all files in a given bucket, which takes very long for the openneuro and openfmri buckets. On my computer, this command (using the example from nipype/interfaces/tests/test_io.py) takes 170 seconds. However, as proposed in my PR, when you use the prefix argument with the bucket_path in the bkt.list() call, it only takes 116 milliseconds. This is because it restricts the filesearch to only the files in the specified bucket_path. The interface still works when bucket_path is not set as input (i.e., it works with the default '' value of the bucket_path parameter).

• Add name/affiliation to zenodo-file (first PR!)