fix(campaign): the action count leaves the setup's login out on a model run

Defects per thousand actions divided by every dispatched step, so a
spec whose setup logs in inflated the denominator by however many steps
that took. It is the same error the run gate had, and it does not cancel
between arms.

A model run is separable because only an llm-selected action stamps
next_action.source. A seeded run is not: its setup returns through the
same entry with no marker, and 11261 dispatched steps across the 169
recorded runs carry no source at all, so excluding on it blind would
report every seeded run as having explored nothing. The seeded arm
counts as before and a test pins that.

cmd/internal-tools/analyze/load.go

@@ -52,10 +52,12 @@ type runRecord struct {
 	// It is what the survival analysis times the event by; see eventStep.
 	FirstViolationDetectedStep *int     `json:"first_violation_detected_step"`
 	ViolatedProperties         []string `json:"violated_properties"`
-	// Actions is the count of steps that dispatched an action, and it is a
-	// pointer so that a runs.jsonl written before the campaign tool counted
-	// them is refused rather than read as an arm that acted zero times. The
-	// campaign tool always emits the field, so its absence dates the file.
+	// Actions is the count of steps on which the action generator dispatched an
+	// action, which excludes the steps the spec's setup drove before the
+	// generator was consulted wherever the trace names them. It is a pointer so
+	// that a runs.jsonl written before the campaign tool counted them is refused
+	// rather than read as an arm that acted zero times. The campaign tool always
+	// emits the field, so its absence dates the file.
 	Actions *int `json:"actions"`
 }