feat: add AI-powered visual regression testing infrastructure

- Add docker-compose.web-testing.yml with vlmkit service
- Add 5 test scripts: capture, compare, pipeline, console-monitor, link-checker
- Add vrt.config.json with 3 viewports + maskSelectors
- Add package.json with @mizchi/vlmkit and Node >=24
- Add tests/README.md with VRT documentation
- Update .gitignore for test artifacts

Based on APAW issue #144 / milestone #106

tests/vrt.config.json

@@ -0,0 +1,19 @@
+{
+  "baseUrl": "http://host.docker.internal:3000",
+  "pages": ["/", "/admin/login"],
+  "viewports": {
+    "mobile": { "width": 375, "height": 667 },
+    "tablet": { "width": 768, "height": 1024 },
+    "desktop": { "width": 1280, "height": 720 }
+  },
+  "maskSelectors": [".timestamp", ".live-counter"],
+  "thresholds": {
+    "pixel": 0.05,
+    "style": 0.02,
+    "a11y": 0.0
+  },
+  "outputDir": "./visual",
+  "ai": {
+    "generateCssFix": false
+  }
+}