Enhancing curation : get_potential_auto_merge()

[yger]

This PR improve get_potential_auto_merge() with:

• several metrics for templates distances
• handling sparsity of templates
• use computed units locations from analyzer

@DradeAW : can you check this, if you use this in Lussac. Normally the behavior is the same as before.

[samuelgarcia]

@alejoe91 this is OK for me. Pierre will need this for circus.

[DradeAW]

I pulled the changed and ran the Lussac unit tests (which check that units that should be merged are indeed merged)

I get this error:

tests/modules/test_merge_units.py:51: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
src/lussac/modules/merge_units.py:56: in run
    potential_merges, extra_outputs = scur.get_potential_auto_merge(self.analyzer, extra_outputs=True, **params['auto_merge_params'])
../../dev/spikeinterface/spikeinterface/src/spikeinterface/curation/auto_merge.py:207: in get_potential_auto_merge
    templates = templates_ext.get_data(outputs="Templates")
../../dev/spikeinterface/spikeinterface/src/spikeinterface/core/sortinganalyzer.py:1704: in get_data
    return self._get_data(*args, **kwargs)
../../dev/spikeinterface/spikeinterface/src/spikeinterface/core/analyzer_extension_core.py:430: in _get_data
    probe=self.sorting_analyzer.get_probe(),
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = SortingAnalyzer: 16 channels - 15 units - 1 segments - memory - has recording
Loaded 2 extensions: random_spikes, templates

    def get_probe(self):
        probegroup = self.get_probegroup()
>       assert len(probegroup.probes) == 1, "There are several probes. Use `get_probegroup()`"
E       AssertionError: There are several probes. Use `get_probegroup()`

../../dev/spikeinterface/spikeinterface/src/spikeinterface/core/sortinganalyzer.py:776: AssertionError

[yger]

Aie. The bug seems to be more in the sorting_analyzer object. Or I should catch the fact that maybe the get_data() for the "templates" extension is not properly able to return Templates when more than one probes are selected. What do you think @samuelgarcia ? Otherwise, if more than one probe is detected, then we get Templates with the 'average' method, as it used to be done before. The following function will work properly since it can handled Templates or numpy arrays as input

[samuelgarcia]

@DradeAW I made a fix for multiple probe.
Can check this again please ?

[DradeAW]

All tests have passed on my end! :)