Fix document analysis: switch to unpdf + mammoth for PDF/Word parsing

pdf-parse v2 requires DOMMatrix (browser API) which fails in Node.js.
Replaced with unpdf (serverless PDF.js build) for PDFs and mammoth for
Word .docx files. Also fixed the same broken pdf-parse usage in
file-content-extractor.ts used by AI filtering.

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

package.json

@@ -68,6 +68,7 @@
     "jspdf-autotable": "^5.0.7",
     "leaflet": "^1.9.4",
     "lucide-react": "^0.563.0",
+    "mammoth": "^1.11.0",
     "minio": "^8.0.2",
     "motion": "^11.15.0",
     "next": "^15.1.0",
@@ -88,6 +89,7 @@
     "sonner": "^2.0.7",
     "superjson": "^2.2.2",
     "tailwind-merge": "^3.4.0",
+    "unpdf": "^1.4.0",
     "use-debounce": "^10.0.4",
     "zod": "^3.24.1"
   },