Reliability improvements: retry logic, validation, ratings sync

- Add 3-attempt retry with exponential backoff to Sreality, Realingo,
  Bezrealitky, and PSN scrapers (CityHome and iDNES already had it)
- Add shared validate_listing() in scraper_stats.py; all 6 scrapers now
  validate GPS bounds, price, area, and required fields before output
- Wire ratings to server /api/ratings on page load (merge with
  localStorage) and save (async POST); ratings now persist across
  browsers and devices
- Namespace JS hash IDs as {source}_{id} to prevent rating collisions
  between listings from different portals with the same numeric ID
- Replace manual Czech diacritic table with unicodedata.normalize()
  in merge_and_map.py for correct deduplication of all edge cases
- Correct README schedule docs: every 4 hours, not twice daily

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

merge_and_map.py

@@ -9,6 +9,7 @@ from __future__ import annotations
 
 import json
 import re
+import unicodedata
 from pathlib import Path
 
 from scrape_and_map import generate_map, format_price
@@ -19,14 +20,8 @@ def normalize_street(locality: str) -> str:
     # "Studentská, Praha 6 - Dejvice" → "studentska"
     # "Rýnská, Praha" → "rynska"
     street = locality.split(",")[0].strip().lower()
-    # Remove diacritics (simple Czech)
-    replacements = {
-        "á": "a", "č": "c", "ď": "d", "é": "e", "ě": "e",
-        "í": "i", "ň": "n", "ó": "o", "ř": "r", "š": "s",
-        "ť": "t", "ú": "u", "ů": "u", "ý": "y", "ž": "z",
-    }
-    for src, dst in replacements.items():
-        street = street.replace(src, dst)
+    # Remove diacritics using Unicode decomposition (handles all Czech characters)
+    street = unicodedata.normalize("NFKD", street).encode("ascii", "ignore").decode("ascii")
     # Remove non-alphanumeric
     street = re.sub(r"[^a-z0-9]", "", street)
     return street