Contract 7.1.2

elixir_timeline after the Elixir Gym's third run on it (feedback #211-#218), with the week-fame fix it shares with the war tools (#223).

• A late capture is a battle the record learned more than a day after it was played, in every window. It had been measured against from, so the same backfilled session was a standout in a wide window and missing from a narrow one around its capture; late captures are counted in battles.late_captures and never narrated (#211).
• A clan entry whose window ends before that week's race closed reads the race as it stood then: fame and place at the last war day closed by to, no race_finished_at, and as_of_window_end: true. A past window had served the week's final 10,146 and a finish five days after to (#213).
• A finished regular week's fame is the banked progress, not the race log's capped 10,000, in week_resolved, the entry's war.resolved and war_history alike: ingest raises it from the day logs and migration 0164 repairs the stored weeks (136/1: 10,146; 130/0: 13,244) (#214, #223).
• A window before the war ledger's start says war moments and war.resolved begin there and points at war_history (#215).
• A past week's clan donations sum the members at the window's end, not today's roster (#216).
• elixir_data_insights says profiles counts every player ever read, including those no longer recorded, beside the recorded population the badge and profile tools draw on (#217).
• The paging findings (#211-#212 continuation cases) are superseded by 7.0.0's newsfeed, which has no continuation page.

Additive; stored week fame and the ledger's week_resolved fame change for weeks the log capped.