reader() is not idempotent

[joshua-gould]

Here's a test that illustrates the unexpected behavior:

 def test_reader_twice(self):
        # calling reader() returns different results
        filename = str(self.path.join("labels"))
        reader = Reader(parse_url(filename))
        assert len(list(reader())) == len(list(reader()))

[will-moore]

Thanks for the feedback.
This seems to be something in the handling of labels.

>>> from ome_zarr.io import parse_url
>>> from ome_zarr.reader import Reader
>>> url = "https://uk1s3.embassy.ebi.ac.uk/idr/zarr/v0.4/idr0062A/6001240.zarr"
>>> reader = Reader(parse_url(url))
>>> list(reader())
[https://uk1s3.embassy.ebi.ac.uk/idr/zarr/v0.4/idr0062A/6001240.zarr/ [zgroup], https://uk1s3.embassy.ebi.ac.uk/idr/zarr/v0.4/idr0062A/6001240.zarr/labels/ [zgroup] (hidden), https://uk1s3.embassy.ebi.ac.uk/idr/zarr/v0.4/idr0062A/6001240.zarr/labels/0/ [zgroup] (hidden)]
>>> list(reader())
[https://uk1s3.embassy.ebi.ac.uk/idr/zarr/v0.4/idr0062A/6001240.zarr/ [zgroup]]

The root node keeps track of nodes is has seen as we traverse the graph, so as to prevent recursion.
The labels nodes are seen on the first call of reader() and this isn't reset for the 2nd call, so the labels nodes are not visited /returned on the 2nd call.

Maybe the API shouldn't allow this and we should just do:

nodes = read(parse_url(url))

or even just:

nodes = read(url)

cc @joshmoore