Harden Facebook Marketplace extraction across DOM variants

[bdtran2002]

Summary

Improve Facebook Marketplace source capture across more real-world DOM shapes and overlays.

Why

The current extractor is heuristic and still has known gaps across public pages, login-gated pages, and layout variants.

Scope

• validate extraction against public, logged-in, and mobile-ish Marketplace layouts
• improve selector fallbacks for title, description, price, seller, and location
• reduce UI chrome leaking into extracted description text
• preserve capture notes when fields cannot be recovered confidently
• add fixture examples for newly supported page shapes

Acceptance criteria

• extractor handles more Marketplace variants with fewer missing-field notes
• description and location extraction are more stable across layout changes
• newly supported variants are captured in reproducible fixtures or documented samples