[eric] fix WebSearch DDG parser, diagnose duplicate text + browser issues

WebSearch fix:
- DuckDuckGo changed HTML attribute order (class before href)
- Added fallback regex to handle both orderings
- WebSearch now returns results correctly

Diagnosed issues (not yet fixed):
- Duplicate text: caused by 2 WebSocket connections to same session
  (frontend React component mounts twice). Pre-existing Haik code issue.
- Browser agent "not found": browser cards require Electron webview,
  doesn't work in regular browser (localhost:3000). Expected behavior.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

backend/apps/agents/tools/web.py

@@ -108,12 +108,19 @@ class WebSearchTool(BaseTool):
             if len(entries) >= num_results:
                 break
 
-            # Title + URL
+            # Title + URL — handle both class-before-href and href-before-class
             link_match = re.search(
                 r'<a[^>]*class="[^"]*result__a[^"]*"[^>]*href="([^"]*)"[^>]*>(.*?)</a>',
                 block,
                 flags=re.DOTALL,
             )
+            if not link_match:
+                # Try reversed attribute order
+                link_match = re.search(
+                    r'<a[^>]*href="([^"]*)"[^>]*class="[^"]*result__a[^"]*"[^>]*>(.*?)</a>',
+                    block,
+                    flags=re.DOTALL,
+                )
             if not link_match:
                 continue