P5 wave 1: extract rotatingMachine concerns into focused modules

src/curves/         loader + normalizer (with cross-pressure anomaly
                      detection) + reverseCurve helper
  src/prediction/     predictors (predictFlow/Power/Ctrl) +
                      groupPredictors (lazy group-scope views) +
                      OperatingPoint (pressure-driven prediction setpoints)
  src/drift/          DriftAssessor (per-metric drift) + PredictionHealth
                      (composes flow/power/pressure into HealthStatus +
                      confidence sibling — see OPEN_QUESTIONS 2026-05-10)
  src/pressure/       VirtualPressureChildren (dashboard-sim) +
                      PressureInitialization (real-vs-virtual tracking) +
                      PressureRouter (dispatches by position)
  src/state/          stateBindings (state.emitter listener helper) +
                      isOperationalState
  src/measurement/    measurementHandlers (dispatcher for flow/power/temp/pressure)
  src/flow/           flowController (handleInput body — execSequence,
                      execMovement, flowMovement, emergencystop)
  src/display/        workingCurves (showWorkingCurves + showCoG admin)
  src/commands/       canonical names: set.mode, cmd.startup/shutdown/estop,
                      set.setpoint, set.flow-setpoint,
                      data.simulate-measurement, query.curves, query.cog,
                      child.register. execSequence demuxes by payload.action
                      to canonical cmd.* handlers.
  CONTRACT.md         inputs/outputs/events/children surface

110 basic tests pass (100 new + 10 pre-existing).
specificClass.js / nodeClass.js untouched — integration in P5 wave 2.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

src/display/workingCurves.js

@@ -0,0 +1,61 @@
+/**
+ * Read-only snapshots of the active machine curves and the centre-of-gravity
+ * statistics. These back the rotatingMachine admin endpoints used by the
+ * editor (`/rotatingMachine/working-curves`, `/rotatingMachine/cog`).
+ *
+ * Both functions accept a single `predictors` argument — an object describing
+ * the current curve state. By taking everything via that one parameter the
+ * helpers stay pure and trivially testable with a plain fixture; the host
+ * just passes itself (or a slim adapter) in.
+ *
+ * Expected shape of `predictors`:
+ * {
+ *   hasCurve: boolean,
+ *   predictFlow, predictPower,        // generalFunctions/predict instances
+ *   getCurrentCurves(): { powerCurve, flowCurve },
+ *   calcCog(): { cog, cogIndex, NCog, minEfficiency },
+ *   cog, cogIndex, NCog,
+ *   minEfficiency,
+ *   currentEfficiencyCurve,
+ *   absDistFromPeak, relDistFromPeak,
+ * }
+ */
+
+const NO_CURVE_ERROR = 'No curve data available';
+
+function showCoG(predictors) {
+  if (!predictors || !predictors.hasCurve) {
+    return { error: NO_CURVE_ERROR, cog: 0, NCog: 0, cogIndex: 0 };
+  }
+  const { cog, cogIndex, NCog, minEfficiency } = predictors.calcCog();
+  return {
+    cog,
+    cogIndex,
+    NCog,
+    NCogPercent: Math.round(NCog * 100 * 100) / 100,
+    minEfficiency,
+    currentEfficiencyCurve: predictors.currentEfficiencyCurve,
+    absDistFromPeak: predictors.absDistFromPeak,
+    relDistFromPeak: predictors.relDistFromPeak,
+  };
+}
+
+function showWorkingCurves(predictors) {
+  if (!predictors || !predictors.hasCurve) {
+    return { error: NO_CURVE_ERROR };
+  }
+  const { powerCurve, flowCurve } = predictors.getCurrentCurves();
+  return {
+    powerCurve,
+    flowCurve,
+    cog: predictors.cog,
+    cogIndex: predictors.cogIndex,
+    NCog: predictors.NCog,
+    minEfficiency: predictors.minEfficiency,
+    currentEfficiencyCurve: predictors.currentEfficiencyCurve,
+    absDistFromPeak: predictors.absDistFromPeak,
+    relDistFromPeak: predictors.relDistFromPeak,
+  };
+}
+
+module.exports = { showWorkingCurves, showCoG };