fix(charts): anchor the Rest chart, which the wrong key left untouched

android/app/src/main/java/com/noop/ui/HealthVitalDetailLogic.kt

@@ -54,7 +54,10 @@ internal fun vo2MaxAttributionSource(estimator: Vo2MaxEstimator?): String =
  */
 internal fun vitalChartYDomain(key: String): ClosedFloatingPointRange<Double>? =
     when (key) {
-        "recovery", "sleep_performance", "strain" -> 0.0..100.0
+        // "rest", NOT "sleep_performance": that is the SERIES name this detail reads underneath, and it is
+        // never a detail key on Android. Writing the series name here compiled, passed a test asserting it,
+        // and left the Rest chart auto-scaling exactly as before.
+        "recovery", "rest", "strain" -> 0.0..100.0
         else -> null
     }
 

android/app/src/test/java/com/noop/ui/VitalChartShapeTest.kt

@@ -20,7 +20,7 @@ class VitalChartShapeTest {
     @Test
     fun `percentage metrics anchor to their natural range`() {
         assertEquals(0.0..100.0, vitalChartYDomain("recovery"))
-        assertEquals(0.0..100.0, vitalChartYDomain("sleep_performance"))
+        assertEquals(0.0..100.0, vitalChartYDomain("rest"))
         assertEquals(0.0..100.0, vitalChartYDomain("strain"))
     }
 
@@ -34,6 +34,27 @@ class VitalChartShapeTest {
         assertEquals(0.0..100.0, vitalChartYDomain("strain"))
     }
 
+    /**
+     * Every key in the allow-list has to be one the screen ACTUALLY receives. The first version anchored
+     * "sleep_performance", which is the series this detail reads underneath and never a detail key here,
+     * so Rest kept auto-scaling while a test asserting that key passed. Pinning against the real key list
+     * is what makes the allow-list checkable rather than plausible.
+     *
+     * `realKeys` is a HAND-MAINTAINED mirror of the `when` in `buildVitalDetail`/`buildSeriesVitalDetail`;
+     * a unit test cannot enumerate a `when`. It catches an anchored key that no screen sends, which is the
+     * bug that shipped. It does NOT catch a newly added metric, so a new key belongs here too.
+     */
+    @Test
+    fun `every anchored key is a real detail key`() {
+        val realKeys = setOf(
+            "recovery", "strain", "resp", "spo2", "rhr", "hrv", "skin", "rest",
+            "fitness_age", "vitality", "vo2max_est", "steps_est",
+        )
+        val anchored = realKeys.filter { vitalChartYDomain(it) != null }
+        assertEquals(setOf("recovery", "rest", "strain"), anchored.toSet())
+        assertNull(vitalChartYDomain("sleep_performance"))   // the series name, not a detail key
+    }
+
     /**
      * Blood oxygen is the case that proves this is not "percentages get 0..100": its real movement lives
      * in 90..100, so anchoring would flatten the signal into a line at the top. Metrics with no fixed