Support registering internal groups/dataset in HDF5.

[danielballan]

A wrapper function takes the same parameters as HDF5Adapter but additionally exposes an optional path, a list of strings navigating to a group or dataset nested inside an HDF5 file.

[danielballan]

Demo of registering a DataSource with shape (1, 5, 5) and then updating that shape to be (2, 5, 5) via PUT /data_source/{path}?data_source={data_source_id}

from tiled.client import from_uri
from tiled.structures.core import StructureFamily
from tiled.structures.data_source import Asset, DataSource, Management
from tiled.structures.array import ArrayStructure, BuiltinDtype
import numpy

c = from_uri('http://localhost:8000', api_key='secret')
arr = numpy.ones((1, 5, 5))
structure = ArrayStructure.from_array(arr)
data_source = DataSource(management='external', mimetype='application/x-hdf5', structure_family='array', structure=structure, assets=[Asset(data_uri='file://localhost/tmp/test.h5', is_directory=False, parameter='data_uri')])
ac = c.new(structure_family='array', data_sources=[data_source])
url = ac.uri.replace('/metadata/', '/data_source/')
ac.refresh()
ds = ac.data_sources()[0]
ds['structure']['shape'] = [2, 5, 5]
ds['structure']['chunks'] = [[dim] for dim in [2, 5, 5]]
ac.context.http_client.put(url, json={'data_source': ds}, params={'data_source': 1})