Real-data application: reprocess δv/v at ≥3 independent public-data stations

[mdenolle]

The persistent reviewer ask and the author's stated next step: select ~3 stations with public continuous data (independent of Okubo 2024 / Kidiwela 2026 / Hotovec-Ellis 2022), measure δv/v, and apply the full forward-and-subtract workflow (§10.4). This also serves as validation for the companion UQ paper.

Context: docs/ai_documentation/09_continuation_handoff_2026-06-25.md (§6 P1).