Spring futureeval 2026 updates

[CodexVeritas]

Summary

• Align spot peer/baseline scoring with Metaculus: N/(N−1) peer factor, include self in the geometric mean, log-scale via zero_point, discrete question support, and peer-baseline filters for non-primary bots / exclude_from_aggregations.
• Fix spot forecast selection to match Metaculus (get_spot_scoring_time chain including open_time, and active window start ≤ spot < end).
• Update dataclip/export SQL for Spring 2026 fields; filter pro exports with is_bot = false and bot exports with is_bot = true.
• Point run_q1_q2_simulations.py at Spring 2026 FutureEval data and keep the cross-tournament control-bot alias list (skip comparison-team outputs if the pair isn’t present).

Validation

• Bot top-30 simulated scores match the Metaculus DB leaderboard exactly.
• Pro scores match within ~0.01.
• Existing unit/integration tests pass (85 passed, 4 skipped).