feat: add caching-logic, extend data-model for shared listings

[chenkel-data]

Summary

This PR extends the data model to support shared listings across multiple agents, improves scrape efficiency with cache-aware early stopping, and updates the UI so the new backend behavior is reflected cleanly in filtering, stats, and listing details.

Main Changes

• Introduce a listing_agents junction table to support many-to-many relationships between listings and agents, allowing a single listing to be associated with multiple agents in overlapping search regions (shared listings).
• Propagate shared-listing tags such as favorites and blacklist status across all linked agents, so actions taken in one agent are reflected everywhere the listing appears.
• Add cache-aware early stopping to scraping, stopping pagination once consecutive pages contain no new listings.
• Update persistence, cleanup, and stats logic to correctly support the shared-listing model.
• Store scrape rank and run metadata to better preserve each provider’s original listing order.
• Update the frontend to support shared listings and show extra context such as listing ID, first seen, and last seen.

Why It Matters

• Listings found in overlapping regions now appear for every relevant agent, instead of being assigned only to the first agent that scraped them.
• The listing order stays closer to the provider’s original order.
• Scraping is now faster by skipping unnecessary pagination for already-known listings.