fix(voice): bound OpenStreetMap text before it reaches voice results

- strip control and format characters from OSM tag values, keeping the
  zero-width (non-)joiners Persian and Indic names spell with, and cut
  them at 120 code points so a name cannot steer or flood the model
- adopt the nine new geometry test files in format-scope.json; five of
  them change by formatting only

src/layers/osmPlaces/index.test.mjs

@@ -17,14 +17,24 @@ function fakeOverlayHost() {
 }
 
 const RECORDS = [
-  { id: 'node/1', lat: 27.7, lon: 85.3, name: 'Bir Hospital', kind: 'hospital' },
+  {
+    id: 'node/1',
+    lat: 27.7,
+    lon: 85.3,
+    name: 'Bir Hospital',
+    kind: 'hospital',
+  },
   { id: 'node/2', lat: 27.71, lon: 85.31, name: null, kind: 'hospital' },
 ];
 
 test('results are queryable only while the layer is on, and labels follow names', () => {
   const host = fakeOverlayHost();
   const layer = createOsmPlacesLayer({ overlayHost: host });
-  layer.setResults(RECORDS, { kind: 'hospital', label: 'hospitals', area: 'Kathmandu' });
+  layer.setResults(RECORDS, {
+    kind: 'hospital',
+    label: 'hospitals',
+    area: 'Kathmandu',
+  });
   assert.deepEqual(layer.getAnalystRecords(), [], 'off: nothing to count');
   assert.equal(layer.getStats().count, 2);
   assert.equal(layer.getStats().statusMessage, 'hospitals · Kathmandu');
@@ -32,10 +42,23 @@ test('results are queryable only while the layer is on, and labels follow names'
   const records = layer.getAnalystRecords();
   assert.equal(records.length, 2);
   records[0].name = 'changed';
-  assert.equal(layer.getAnalystRecords()[0].name, 'Bir Hospital', 'callers get copies');
-  assert.deepEqual(layer.getResultsMeta(), { kind: 'hospital', label: 'hospitals', area: 'Kathmandu', count: 2 });
+  assert.equal(
+    layer.getAnalystRecords()[0].name,
+    'Bir Hospital',
+    'callers get copies',
+  );
+  assert.deepEqual(layer.getResultsMeta(), {
+    kind: 'hospital',
+    label: 'hospitals',
+    area: 'Kathmandu',
+    count: 2,
+  });
   layer.disable();
-  assert.ok(host.calls.some(([name, id]) => name === 'clearSource' && id === 'osm-places'));
+  assert.ok(
+    host.calls.some(
+      ([name, id]) => name === 'clearSource' && id === 'osm-places',
+    ),
+  );
 });
 
 test('an empty layer tells the operator how to fill it', () => {

src/layers/osmPlaces/model.js

@@ -8,8 +8,33 @@ export const OSM_PLACES_LAYER_ID = 'osm-places';
 /** A node and a way/relation of the same name this close are one site mapped twice. */
 const SAME_SITE_M = 75;
 
-const text = (value) =>
-  typeof value === 'string' && value.trim() ? value.trim() : null;
+/** Longest tag value kept; OSM names are open text that reaches the model. */
+const MAX_TEXT = 120;
+
+/**
+ * A tag value as plain bounded text: control characters (newlines, tabs)
+ * become spaces; format characters (bidi overrides, zero-width spaces) are
+ * dropped, except the zero-width (non-)joiners Persian and Indic names spell
+ * with; runs of whitespace collapse, and long values are cut with an
+ * ellipsis.
+ */
+function text(value) {
+  if (typeof value !== 'string') return null;
+  const clean = value
+    .replace(/\p{Cc}+/gu, ' ')
+    .replace(/(?![\u200c\u200d])\p{Cf}/gu, '')
+    .replace(/\s+/g, ' ')
+    .trim();
+  if (!clean) return null;
+  // Cut on code points so an emoji's surrogate pair is never split.
+  const chars = Array.from(clean);
+  return chars.length <= MAX_TEXT
+    ? clean
+    : `${chars
+        .slice(0, MAX_TEXT - 1)
+        .join('')
+        .trimEnd()}…`;
+}
 
 function distanceM(a, b) {
   const toRad = Math.PI / 180;