Add automatic post-sorting unit merge (Phase 3b) via SpikeInterface curation

[Copilot]

Kilosort4 frequently oversplits single neurons into multiple clusters. This adds a dedicated auto-merge phase between the analyzer and report generation that detects and corrects those splits using SpikeInterface's auto_merge_units.

Pipeline change

New Phase 3b inserted between run_analyzer() and generate_reports():

pipeline.run_sorting()
pipeline.run_analyzer()
pipeline.run_auto_merge(no_auto_merge=resolved["no_auto_merge"])  # new
pipeline.generate_reports(...)

run_auto_merge() calls sic.auto_merge_units() with the "similarity_correlograms" preset, which gates merges through six sequential criteria:

1. Minimum spike count (preset default: 100)
2. Refractory-period contamination filter
3. Location proximity (≤150 µm)
4. Template shape similarity
5. Cross-correlogram similarity
6. Net quality-score improvement (firing rate vs. contamination balance)

Merged units are saved to analyzer_merged/ and loaded on resume. If no merge candidates are found, or if the step throws, the original analyzer is used unchanged.

New controls

Flag	Config key	Effect
--no-auto-merge	curation.no_auto_merge	Skip the merge step entirely

Available on both mea_analysis_routine.py and run_pipeline_driver.py; propagated through build_extra_args.

Docs

README.md updated with: pipeline stages table, key features list, output directory layout (analyzer_merged/), config section reference, CLI reference table, workflow example, and a new Automatic Unit Merging section covering the algorithm and disable instructions.

---

💬 Send tasks to Copilot coding agent from Slack and Teams to turn conversations into code. Copilot posts an update in your thread when it's finished.

[mandarmp]

@mandarmp @adamwea I had started this experimentation with the automerge feature. Is it similar to yours?

[adamwea]

@mandarmp @adamwea I had started this experimentation with the automerge feature. Is it similar to yours?

Yea I have automerge implemented on my branch as an alternative to unit match. Just the same way it's implemented in si docs.

[mandarmp]

@adamwea can we have your PR then we can add that functionality here to the MAIN BRANCH