test(voice): record the measured prompt size, 55,210

The budget test measures the full instructions plus every tool. At this
head that is 55,210, ~290 under the 55,500 ceiling; the note said 55,370,
which that measurement does not produce.

src/voice/pointAndAsk.test.mjs

@@ -345,10 +345,10 @@ test('deixis prompt lint: one short rule, and every sentinel it names exists in
   // policy once (WHERE precedence, AREAS routing, AFTER TOOLS speech) brought
   // it to 51,646. Voice geometry (resolve_area, find_imagery, osm_query and
   // the analyst area scopes) raised the ceiling to 55,500; tightening the
-  // AREAS rule and the area tool descriptions left it at 55,263 with an
-  // operator Overpass; main's Sea Marks and Ski Pistes layers bring it to
-  // 55,370, ~130 under. A new tool or rule
-  // has to fit under this ceiling.
+  // AREAS rule and the area tool descriptions left it at 55,103 as measured
+  // below (every tool, osm_query included); main's Sea Marks and Ski Pistes
+  // layers bring it to 55,210, ~290 under. A new tool or rule has to fit
+  // under this ceiling.
   const size = instructions.length + JSON.stringify(GEV_REALTIME_TOOLS).length;
   assert.ok(size <= 55_500, `prompt size ${size}`);
 });