ndi.app.spikesorter and ndi.app.spikeextractor are frameworks, and are not tracked by #97 or anywhere else #135

Part of #90 (PR #61 catch-up). Surfaced while checking the triage correction: these two were not in #90's list at all, and I could not find them named in any other open issue.

Not covered by #97

#97 is the nearest issue, but reading it, it covers:

• the Kilosort import pipeline (fun/probe/import_/kilosort/*) — since landed on main via Port ensembleMaker and spikeSorterImporter, with the kilosort import subsystem beneath it #125;
• the interactive sorter GUI: src/ndi/app/spikesorter_gui.py and spikesorter_clustermodel.py;
• the .vsw and KlustaKwik file formats.

The headless apps — ndi.app.spikesorter and ndi.app.spikeextractor — are named nowhere in it. They are the layer the GUI drives, and they are useful without a GUI.

State on main

Stubbed methods:

• spikeextractor: makefilterstruct, filter, extract
• spikesorter: loadwaveforms, spike_sort, clusters2neurons

Ordering

spikeextractor first: it is the upstream half (filter the raw trace, extract waveforms), it has no clustering dependency, and spikesorter.loadwaveforms reads what it writes.

Two things to settle before starting:

• .vsw is likely a prerequisite for MATLAB-compatible storage of extracted waveforms, and .vsw is in Kilosort import pipeline, interactive spike-sorter GUI, and the .vsw / KlustaKwik file formats #97. Either sequence this behind that part of Kilosort import pipeline, interactive spike-sorter GUI, and the .vsw / KlustaKwik file formats #97, or establish that the Python-side storage can land first and be made MATLAB-readable afterwards. Worth deciding explicitly rather than discovering mid-port.
• clusters2neurons creates ndi.neuron elements — which is exactly the class broken by ndi.neuron and ndi.element.timeseries report the wrong ndi_element_class and are unregistered: neurons are silently lost in both directions #133 (wrong ndi_element_class, unregistered). Landing this before ndi.neuron and ndi.element.timeseries report the wrong ndi_element_class and are unregistered: neurons are silently lost in both directions #133 means writing neurons that read back as plain elements. ndi.neuron and ndi.element.timeseries report the wrong ndi_element_class and are unregistered: neurons are silently lost in both directions #133 should go first.

Notes

• As with ndi.app.oridirtuning is a framework: all five compute methods raise NotImplementedError #134, the branch's version predates the DID rewrite and the ndi.fun.utils.identifier convention; treat it as a reference, not a merge candidate.
• Check the appdoc document/schema pairs are in ndi_common before starting (cf. ndi_common re-sync from NDI-matlab: probe-geometry library, missing document/schema pairs, and the probe_geometry matrix invariant #96).
• src/ndi/app/ndi_matlab_python_bridge.yaml needs an INTERFACE UPDATE when this lands.