i18n: fix the LGTM review findings and tidy the branch texts

- i18n scanner: lastError sink, no false sinks on defaults and ternaries
- shared src/i18n/marker.js for modules that only mark strings
- wind field messages move to wind/fields.js; CITY_POIS imports from cityPois.js
- drop review IDs, change history and stale references from comments and docs

Co-Authored-By: Claude Opus 5.5 <[email protected]>

src/voice/voiceCost.js

@@ -50,7 +50,7 @@ export const VOICE_MODELS = Object.freeze({
   standard: Object.freeze({
     tier: 'standard',
     id: 'gpt-realtime-2',
-    label: 'STANDARD', // i18n-ignore: tier name for reports and scripts; the app's tier button sets its own text
+    label: 'STANDARD', // i18n-ignore: model data; the tier button has its own text
     /** USD per 1M tokens — gpt-realtime-2. */
     rates: Object.freeze({
       textInput: 4,
@@ -66,7 +66,7 @@ export const VOICE_MODELS = Object.freeze({
   mini: Object.freeze({
     tier: 'mini',
     id: 'gpt-realtime-2.1-mini',
-    label: 'MINI', // i18n-ignore: tier name for reports and scripts; the app's tier button sets its own text
+    label: 'MINI', // i18n-ignore: model data; the tier button has its own text
     /** USD per 1M tokens — gpt-realtime-2.1-mini (~3.2× cheaper on audio). */
     rates: Object.freeze({
       textInput: 0.6,