Shuffling data in HDF5Pump

Summary

@mmoser would like to have the ability to shuffle groups when reading an HDF5 file with HDF5Pump.

Something like this:

pipe.attach(HDF5Pump, filename='...', shuffle=True)

What is the current workaround to achieve the same functionality?

Reading the group_ids from the /group_info table in advance, shuffle that and create a one-to-one-map and then change the group_id of each blob item with a separate module.