write Blob returned by Pipeline.drain with io.HDF5Sink

Summary

The need for the following feature request is meant to be discussed first, of course. Maybe it's not of general interest or I'm misusing the classes anyway.

Draining a pipeline returns a Blob, which is filled by the returned value of the pre_finish method of any attached module.

Example

class Finalizor(kp.Module):
    def pre_finish(self):
        return kp.Table({ 'useless': np.zeros(5) }, h5loc='/finally');

class Nihil(kp.Module):
    def process(self, blob):
        raise StopIteration    

pipe=kp.Pipeline()
pipe.attach(Finalizor)
pipe.attach(Nihil)
final_blob = pipe.drain()

sink = kp.io.HDF5Sink(filename = 'test.h5')
for mod in final_blob:
    if isinstance(final_blob[mod], kp.dataclasses.Table):
        sink.write_table(final_blob[mod])
sink.finish()

Note that final_blob is

Blob([('Finalizor',
       Generic Table <class 'km3pipe.dataclasses.Table'> (rows: 5)),
      ('Nihil', None)])

and contains None for the attached Nihil module, which cannot be wrote to disk, of course.

Feature request

The loop above to check the final_blob on existing kp.Tables could be implemented into a new service HDF5Sink.write_blob. In case the blob contains nested data, i.e. further Blobs with tables, write_blob could recursively iterate through the final_blob and write the tables to disk.

What do you think?