proxywatchd: mark confirmed-dead proxies as permanently dead

- Add DEAD_PROXY=-1 constant for permanently dead proxies
- Mark proxy dead when: failed >= max_fail*2, or max_fail with fatal error
- Fatal errors: refused, unreachable, auth (proxy definitely not working)
- Dead proxies excluded from testing (failed >= 0 query)
- Cleanup_stale also removes old dead proxies
- Dashboard shows separate dead vs failing counts

httpd.py

@@ -126,11 +126,17 @@ def get_db_health(db):
         ).fetchone()
         stats['added_last_day'] = row[0] if row else 0
 
-        # Dead proxies count
+        # Dead proxies count (permanently dead = -1, failing = positive)
+        row = db.execute(
+            'SELECT COUNT(*) FROM proxylist WHERE failed = -1'
+        ).fetchone()
+        stats['dead_count'] = row[0] if row else 0
+
+        # Failing proxies count (positive fail count but not permanently dead)
         row = db.execute(
             'SELECT COUNT(*) FROM proxylist WHERE failed > 0'
         ).fetchone()
-        stats['dead_count'] = row[0] if row else 0
+        stats['failing_count'] = row[0] if row else 0
 
     except Exception:
         pass