feat: Replace DuckDuckGo with direct site search

- Remove duckduckgo-search dependency (unreliable results)
- Add direct_search_service.py that scrapes sites' search pages directly
- Add search_url and product_link_selector fields to SearchSite model
- Add beautifulsoup4 and lxml for HTML parsing
- Update schemas with new fields
- Add database migration for new columns

Direct search is more reliable as it queries each site's own search
functionality instead of relying on external search engines.

pyproject.toml

@@ -23,7 +23,8 @@ dependencies = [
     "alembic",
     "tenacity",
     "slowapi",
-    "duckduckgo-search",
+    "beautifulsoup4",
+    "lxml",
 ]
 
 [build-system]