Rewrite Niedersachsen spider with shapefile-first approach

Replaces API-only scraping with LSN geodata integration providing
100% geolocation coverage (4,250 schools). Implements robust matching,
normalization, and async downloads.

Key improvements:
- Add shapefile integration for 4 LSN categories (ABS, Förder, BBS, SdG)
- Implement Unicode-aware (NFKD) German character normalization
- Add collision-safe API indexing with form-aware disambiguation
- Use async Scrapy Requests for shapefile downloads (non-blocking)
- Generate stable synthetic IDs (SHA-1) for shapefile-only schools
- Add path traversal protection in ZIP extraction
- Ensure API-only schools are fetched after shapefile processing
- Add enhanced school form normalization with long-form variants
- Add allowed_domains and stats tracking

Coverage: 4,250 schools with geodata (~75% API-enriched, ~25% shapefile-only)

Adds pyproj dependency for CRS transformation support.

Note: API endpoint (schulen.nibis.de) is public but may experience
intermittent connectivity. Shapefile processing is fully functional
and provides complete geodata coverage.

Author: tim

README.md

@@ -49,9 +49,9 @@ When available, we try to use the geolocations provided by the data publishers.
 | BB    | ✅ Yes                | WFS                                          |
 | HB    | ❌ No                 | -                                            |
 | HH    | ✅ Yes                | WFS                                          |
-| HE    | ❌ No                 | -                                            |
+| HE    | ⚠️  Partial (90.7%)   | Extracted from OSM on detail pages (1,863/2,054 schools). The 191 schools without coordinates include both schools with placeholder coordinates (-1.0, -1.0) that are filtered to null and schools with no map data at all. |
 | MV    | ✅ Yes                | WFS                                          |
-| NI    | ❌ No                 | -                                            |
+| NI    | ✅ Yes (4,250 schools) | Shapefile-first approach: all 4,250 LSN schools have geodata. ~75% enriched with API data (phone, email, etc.), ~25% basic shapefile data only. Requires manual shapefile setup in cache/niedersachsen_shapefiles/. |
 | NW    | ✅ Yes                | Converted from EPSG:25832 in source CSV data |
 | RP    | ❌ No                 | -                                            |
 | SL    | ✅ Yes                | WFS                                          |