Fix spatial attribute positions of sliced NodeCollection

[hakonsbm]

For internal consistency and efficiency, metadata of nodes in a sliced NodeCollection points to the metadata of the original NodeCollection. When getting the positions in the spatial parameter of a sliced NodeCollection, using nodes.spatial['positions'], positions for all nodes in the original NodeCollection is therefore returned. This PR adds a filter on the returned positions to only get positions of nodes in the sliced NodeCollection, which fixes #2414.

[mlober]

The test only creates a single neuron when it should probably create multiple neurons and then create a sliced NC from that.

[hakonsbm]

@mlober @nicolossus Thanks for your reviews. I've updated with the suggested changes. Please have another look.

[nicolossus]

Thanks for the changes!

[mlober]

Thanks for the changes. I still think the node collection object should be sliced first for the test, since now the NC is a standard one with size 10.
I think it would work if you added the line
nodes = nodes[::2] in the code.
If I am misunderstanding, please explain.

[hakonsbm]

@mlober The test already checks positions of a NC sliced to a single node. The positions of each node are gathered to a list that can be compared to the original positions:

nest-simulator/testsuite/pytests/test_spatial/test_basics.py

Line 429 in c99d3c5

all_positions = sum([list(nodes[i].spatial['positions']) for i in range(len(nodes))], start=[])

However, I've added a check of positions sliced with nodes[::2] as well now.

[mlober]

I see, thanks!