NWB recording with multiple electrode groups (e.g. probes) has non-unique channel locations

Right now when spikeinterface opens an NWB file as a si.Recording, the channel locations are defined by looking up rel_x etc columns in the electrodes table of the NWB file. But because rel_x etc is technically the relative coordinates within an electrode group, it can lead to non-unique channel locations. This will happen e.g. if the NWB file has electrical series from two probes that are of identical type but implanted in different locations and the electrode groups are defined at the level of probes. This seems bad to me but not sure what the solution is. Could use the x etc columns in electrode table, or looking up target locations that may be a property of electrode groups. Alternatively could return a list of recordings, one for each electrode group, rather than a single recording objects.

related discussion: https://github.com/LorenFrankLab/trodes_to_nwb/issues/120