Loading channel locations in nwb recordings

[khl02007]

@lfrank @magland @alejoe91

summary: #552 didn't resolve our issue

issue: we store data from both tetrodes and probes in the same NWB file. The location of the channels in a tetrode is defined on the xz plane (e.g. (rel_x, rel_y, rel_z) = (6.25, 0.0, 6.25)). On the other hand, the probe channels are defined on the xy plane ((rel_x, rel_y, rel_z) =(1086.00, -1080.0, 0.00)). right now, when we load the nwb file as a recording object in spikeinterface, the tetrode channel locations are not returned correctly even after specifying the axes. For example, for a tetrode channel 0, recording.get_channel_locations(channel_ids=[0], axes='xyz') returns array([[6.25, 0. , 0. ]]) rather than array([[6.25, 0. , 6.25 ]]).

I think in the future spikeinterface should move toward 3d definition of channel locations as @magland and others suggested.