Admin: Show chart game-version history separately from dataset publications

[gekichumai-admin]

Phase / step: Phase 4 — Provenance / Step 4
Parent: #301
Depends on: #335, #330

Context and goal

A chart's game-version evolution is a different concept from the system's dataset-publication revisions. The UI must preserve that distinction while showing what changed between game versions and the evidence/reason behind the change.

Scope and implementation notes

• Add a clearly labelled game-version history view alongside, but not merged into, the dataset-publication timeline.
• Show each known game-version state in domain order with effective version/range, observed values, change summary, decision reason, origin/evidence, and completeness status.
• Compare each state with the preceding game-version state and allow a field-level expansion for additions, removals, and changes.
• Link a game-version state to the dataset publication(s) that first recorded or later revised that knowledge without implying that the two revision axes are equivalent.
• Mark backfilled, unknown, and conflicting historical evidence explicitly.
• Provide durable route parameters for chart and selected game-version transition so reports, comments, and dashboard items can link to it.
• Use domain labels and explanatory copy to prevent users from confusing “game version” with deployment version or dataset revision.

Acceptance criteria

• Game-version history and dataset-publication history are separate, clearly named views.
• Each game-version entry includes its effective version, relevant changes, reason, evidence origin, and completeness state.
• Field-level changes from the prior game-version state can be inspected.
• Cross-links to related dataset revisions preserve both identities in the destination URL.
• Backfilled, unknown, and conflicting evidence is visible and filterable where supported.
• Deep links restore the selected game-version transition.
• Empty and single-version histories have useful explanatory states.
• Tests assert that dataset revision labels are never presented as game versions.

Out of scope

• Editing game-version assignments.
• Predicting future game-version changes.
• Replacing the arbitrary dataset-revision comparison.
• Pipeline execution or rollback.