Stop truncating hypothesis description to 700 chars in filter_score

filter_score.js line 116 builds the userPrompt with `${h.description.substring(0, 700)}` — descriptions in the hypotheses table routinely exceed 1200 chars (the genesis spec asks for 120-200 words = ~1000-1500 chars and many proposals run to 2000+ after PIVOT revisions). The current truncation cuts off the structural_veto, resolution_event, archetype_id, and avoided_traps that the genesis proposer added — exactly the fields the LOW/HIGH reviewers should be scoring against. Change line 116 to substring(0, 2500). Token cost delta: ~+800 input tokens × 2 calls × 3 runs = ~5K extra input tokens per hypothesis (~$0.005 on Opus, negligible on Gemini).

On 20 hypotheses with description > 700 chars, rerun filter_score with the 2500-char prompt and compare composite_rank_score against the historical 700-char score. Expect median absolute delta of 0.5-1.5 on composite; the solo_founder_feasible axis specifically should shift on hypotheses whose distribution surface was named only in chars 700-2000.

If median absolute composite delta across the 20 hypotheses is < 0.2, the additional context is not changing scoring — the truncation was harmless. Revert.

Rejected at the council verdict. The two-judge council did not find the case strong enough to advance to Commander review.