perf: geocode pipeline — batch POST + parallel server, 200→1 roundtrip

JS preload(): replaced N individual GET /api/geocode (concurrency 2) with
a single POST /api/geocode/batch — one HTTP roundtrip instead of up to
N/2 sequential ones. schedLoc debounce reduced 600→200ms.

Server geocode_batch(): ThreadPoolExecutor(max_workers=4) parallelises
Nominatim/reverse_geocoder calls instead of sequential loop. Deduplicates
coords before submitting futures. lru_cache absorbs duplicates within batch.

Pi geocode_batch_route(): deduplicate coords, removed stale 'if label:'
guard (label always non-empty after BUG-13 fix).

app.py

@@ -212,6 +212,9 @@ def geocode_batch_route():
     coords = body.get("coords")
     if not isinstance(coords, list) or not coords:
         return jsonify(error="expected {coords: [{lat, lon}, ...]}"), 400
+    # Deduplicate — same (lat4,lon4) would hit lru_cache anyway but avoids
+    # submitting redundant work.
+    seen = set()
     result = {}
     for item in coords:
         try:
@@ -221,9 +224,9 @@ def geocode_batch_route():
             continue
         lat4 = int(round(lat, 4) * 10000)
         lon4 = int(round(lon, 4) * 10000)
-        label = _geocode_cached(lat4, lon4)
-        if label:
-            result[f"{lat:.3f},{lon:.3f}"] = label
+        if (lat4, lon4) not in seen:
+            seen.add((lat4, lon4))
+            result[f"{lat:.3f},{lon:.3f}"] = _geocode_cached(lat4, lon4)
     return jsonify(result)
 
 @app.route('/api/dash/devices')