Add post-spike repolarization (fall-time) test to all neuron presets

[JCorson]

Background

While reviewing PR #318 (DA pacemaking fix for #304), an AP-shape pathology was found in the SNc Dopaminergic preset that no existing test caught: the cell would spike normally (peak ~+27 mV, half-width ~1.7 ms — passing all bands), but then park at −30 mV for 20+ ms after the peak before finally dropping to AHP. Root cause was the Komendantov Na window current (m_inf = 0.79 at −30 mV) dominating the post-spike conductance landscape; SK / Ca dynamics were retuned in #318 to fix it.

Crucially, mean_half_width does not catch this. Half-width is measured at the spike midpoint (~−10 mV) — V can cross that level cleanly during the upstroke and downstroke and still hang at −30 mV afterwards. Likewise mean_ahp_depth finds the eventual AHP minimum even if it arrives 30 ms late.

The new regression test added in #318 — test_da_single_ap_repolarises_cleanly in tests/integration/test_dopaminergic.py — drives a single AP with the UI ACTION_POTENTIAL protocol (4 µA/cm² × 5 ms) and asserts that V drops below −60 mV within 5 ms after the peak. This pins the post-spike repolarization shape, which is what was actually broken.

Proposal

Add an analogous fall-time test to the preset test files for every preset in NEURON_PRESET_NAMES:

• tests/integration/test_squid_giant_axon.py
• tests/integration/test_cortical_pyramidal.py
• tests/integration/test_fast_spiking_interneuron.py
• tests/integration/test_purkinje.py
• tests/integration/test_thalamic_relay.py
• tests/integration/test_ca1_pyramidal.py
• tests/integration/test_stn.py
• tests/integration/test_trn.py

(plus any species-specific files I've missed)

Each test should:

1. Drive a single evoked AP with that preset's ACTION_POTENTIAL protocol parameters.
2. Assert V drops below a species-appropriate threshold (typically −55 to −65 mV) within a species-appropriate window after the peak.

The exact threshold/window pair must come from the species literature. Examples:

• Fast-spiking interneuron: very fast — V should drop below −60 mV within 1–2 ms after peak.
• Cortical pyramidal: standard — within ~3–5 ms.
• Thalamic-relay (tonic mode only): tonic AP only; the burst LTS plateau is a separate phenotype and should not be exercised by this test.
• Purkinje: fast — peak to AHP within 2–3 ms.
• CA1 pyramidal: ~4–6 ms.
• STN: ~3–5 ms.
• TRN (tonic mode only): same caveat as TC — burst-mode rebound LTS is separate.
• Squid giant axon: HH52 reference; ~5 ms.

For presets that have both tonic and burst modes (TC, TRN), this test should target tonic mode explicitly, since the burst/LTS plateau is a real biological phenomenon and must not be conflated with a pathological Na-window plateau.

Why this matters

Without this test:

• Half-width and AHP-depth bands can pass while the AP shape is fundamentally wrong (fix(presets): DA SNc pacemaking and AP shape via INaP + Mainen-Sejnowski Kv #318 review).
• F-I curves can look pathologically wrong because the post-spike plateau prevents proper inter-spike repolarization, but the existing tests only check spike count and rolling-mean voltage — they miss shape problems.
• A future change to a Na/K kinetic or a leak conductance can silently break repolarization in any preset.

Acceptance criteria

• A test_<species>_single_ap_repolarises_cleanly test in each preset's test file.
• Each test cites the literature source for its threshold/window choice.
• All tests pass on main.
• Where literature does not support a clean fall-time bound (e.g. burst-prone presets), the test is either restricted to tonic mode or explicitly omitted with a comment explaining why.

References

• Discovered in PR fix(presets): DA SNc pacemaking and AP shape via INaP + Mainen-Sejnowski Kv #318 review.
• Root-cause discussion: search "Na window current" / "post-spike plateau" in PR fix(presets): DA SNc pacemaking and AP shape via INaP + Mainen-Sejnowski Kv #318 review thread.
• Reference test: tests/integration/test_dopaminergic.py::test_da_single_ap_repolarises_cleanly.