Throughstone v1.7.1 — inputs lifecycle

A small follow-up that gives the inputs/ folder (added in 1.7.0) a lifecycle, so brought-in documents don't quietly go stale as the generated architecture/ docs supersede them. Greenfield-inert: a fresh project ships an empty ledger and behaves exactly as before.

Added

• Inputs index (Code/{{PROJECT}}-docs/inputs/inputs-index.md) — a ledger that records, per input, which parts a generated architecture/ / adr/ doc has superseded and which still hold (section-level where the input's structure allows, whole-doc otherwise). It is written when an input is imported and when a session captures its content.
• inputs/archive/ retirement — a fully-superseded input is moved there (a move, never an edit or a delete); sessions read inputs/ but not inputs/archive/, so a captured document stops reading as current intent while it is kept for history.
• Check-in inputs sweep — the periodic check-in reconciles the ledger against the architecture docs, surfaces a newly-superseded input for a retire/keep decision (never auto-moving), and records the outcome in the check-in report.

Changed

• Inputs are point-in-time; architecture/ is the living truth. Where a generated architecture/ or adr/ doc covers the same ground as an input, the generated doc wins. Architecture-grade inputs are lifted into architecture/ promptly — a PRD synthesized, a protocol/API spec or other finished doc copied or lightly reformatted — rather than left living in inputs/. The system-overview, architecture-overview, substep, and implementation-planning reminders now read live inputs (not inputs/archive/) and treat them as a starting point, not current truth.

Migration: see UPDATING-THROUGHSTONE.md → "1.7.1 migration".

Full changelog: v1.7.0...v1.7.1