Make scraper resilient to Google Maps DOM changes (selector fallbacks, consent handling, geo data)

[Rmaawn]

Summary

The scraper relied on hard-coded XPaths tied to Google's obfuscated CSS classes
(e.g. TIHn2, DUwDvf lfPIob). Google rotates these frequently, which silently
breaks extraction. This PR reworks the core to be resilient and adds several
commonly-needed fields.

What changed

• Selector fallback layer — new safe_text / safe_attr helpers try a
  prioritized list of CSS + XPath selectors and return the first valid match,
  so a single class change no longer breaks a field.
• Consent / GDPR handling — pre-seeds CONSENT/SOCS cookies and adds a
  dismiss_consent fallback, preventing the scraper from getting stuck on
  consent.google.com.
• Direct search-URL navigation — navigates straight to /maps/search/...
  instead of typing into the search box (more stable, fewer race conditions).
• Smarter result loading — scroll_results detects the end-of-list marker
  and stops on stale rounds instead of fixed sleeps.
• New fields — latitude, longitude, plus_code, and google_maps_url
  parsed from the place URL.
• Single-result handling — detects and extracts a direct one-place hit.
• QoL — --headless flag and utf-8-sig CSV output (correct Unicode in Excel).

Notes

• Backward compatible with the existing CLI flags; no breaking changes.
• DOM selectors will still need occasional updates as Google evolves, but the
  fallback layer makes failures degrade gracefully instead of breaking outright.