Harden watch sync against schema mismatches and surface log-row settings

The watch now re-applies the phone's application context once activation
completes (real hardware activates asynchronously, so the eager launch
read sees an empty context), and a state push that fails to decode —
a phone/watch build running different document schemas — is logged and
skipped instead of pruning the cache against a bogus empty set. Workout
log rows offer the machine-settings editor for any machine-based library
exercise, not just logs that already carry settings.

Claude-Session: https://claude.ai/code/session_01LEoff8bXGBS83tK1c55Mf7

Workouts/Views/WorkoutLogs/WorkoutLogListView.swift

@@ -434,11 +434,14 @@ private struct WorkoutLogRow: View {
                         .foregroundStyle(.primary)
                 }
 
-                // Machine comfort settings — shown for any machine-based log
-                // (`machineSettings != nil`), independent of load type. Tapping opens
-                // the editor sheet; `.plain` keeps the tap from firing the row's
-                // NavigationLink (same trick as the checkbox button).
-                if log.machineSettings != nil {
+                // Machine comfort settings — shown when the log carries settings, or
+                // for any exercise the authored library identifies as machine-based
+                // (same rule as ExerciseView), so older logs offer the affordance
+                // before settings are first recorded. Tapping opens the editor sheet;
+                // `.plain` keeps the tap from firing the row's NavigationLink (same
+                // trick as the checkbox button).
+                if log.machineSettings != nil
+                    || ExerciseInfoLibrary.info(for: log.exerciseName)?.isMachineBased == true {
                     Button(action: onSettingsTap) {
                         Label("Settings", systemImage: "slider.horizontal.3")
                             .font(.footnote.weight(.medium))