Enforce lossless Prefill token budgets

[FluffyAIcode]

Summary

• cap Strategy Prefill at 4096 tokens using complete active-leaf ancestry and exact relevant experiment records
• cap Generator/Critic Prefill at 6144 tokens with preflight rejection instead of truncation
• remove results tail slicing and prose-plan step truncation while preserving full Critic context

Test plan

• targeted AutoResearch tests: 41 passed
• compile supervisor and GAN REPL
• verify over-budget inputs remain unchanged and are rejected before remote Prefill

The uncommitted candidate.py file is active runtime state and is excluded.

Made with Cursor

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.