feat: assessed filter, 5000 per-page limit, auto-advance on empty Not-checked page

Assessed/Not assessed filter:
- 'yes' → beauty_lead_quality IS NOT NULL (has been B2B assessed)
- 'no'  → beauty_lead_quality IS NULL (never assessed)
- wired through /api/enriched → get_enriched(beauty_assessed=)

Per-page limit:
- options: 100 / 500 / 1000 / 2000 / 5000
- backend cap raised from le=1000 to le=5000

Auto-advance on empty Not-checked page:
- after bulk validate/prescreen, loadDomains reloads the same DuckDB page
- if every domain on that page is now processed (client-side filter → 0 rows)
  but the page still returned results, automatically increment page and retry
- prevents "No domains found" after successfully processing a batch
- capped at page 500 to avoid infinite loop

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

app/beauty_main.py

@@ -157,14 +157,16 @@ async def enriched(
     tld: str = Query(None),
     alpha_only: bool = Query(False),
     no_sld: bool = Query(False),
+    assessed: str = Query(None),
     page: int = Query(1, ge=1),
-    limit: int = Query(100, ge=1, le=1000),
+    limit: int = Query(100, ge=1, le=5000),
 ):
     total, rows = await get_enriched(
         min_score=min_score, country=country,
         prescreen_status=prescreen_status, niche=niche, site_type=site_type,
         keyword=keyword, tld=tld,
         alpha_only=alpha_only, no_sld=no_sld,
+        beauty_assessed=assessed,
         page=page, limit=limit,
     )
     return {"page": page, "limit": limit, "total": total, "results": rows}