Add normalized-result v2 execution identity

[chenmingtang830]

Summary:

• Introduce ax.normalized-result/v2 with product, surface, harness, model, and effort identity while retaining unchanged v1 compatibility.
• Split producers, records diff, arena selection, and publication exports by model and effort.
• Reject trial aggregation across execution identity or evaluation-set boundaries.
• Show model and effort in reviewer-facing diffs and competitive reports.

Migration:

• New aggregate records emit v2.
• records-diff continues to read v1; missing effort is shown as unknown.
• ax.normalized-cell-record/v1 remains unchanged.

Verification:

• npm run typecheck
• npm test: core 546, arena 285
• npm run build
• npm pack dry run

Stacked on #219.

[github-actions]

Normalized records diff

✅ PASS — No correctness regression detected

Compared 4 baseline surface cell(s); no existing pass@1 or pass³ score decreased or disappeared.

Overall is a macro-average of participating surface scores. Harness/model/effort configurations are never averaged together.

Overall by execution configuration

Product	Harness	Model	Effort	Base pass@1	Head pass@1	Δ	Base pass³	Head pass³	Surfaces
fixture-db	claude-code	claude-sonnet-5	—	92.9%	92.9%	+0.0 pp	78.6% (11/14)	78.6% (11/14)	2
fixture-db	codex	gpt-5.6-terra	—	78.6%	78.6%	+0.0 pp	64.3% (9/14)	64.3% (9/14)	2

Surface cells

Product	Harness	Model	Effort	Surface	Base pass@1	Head pass@1	Δ	Base pass³	Head pass³	Latency ms	Cost USD	Tokens	Harness version
fixture-db	claude-code	claude-sonnet-5	—	api	85.7%	85.7%	+0.0 pp	71.4% (5/7)	71.4% (5/7)	1100 → 1100	0.4200 → 0.4200	3350 → 3350	2.1.5 → 2.1.5
fixture-db	claude-code	claude-sonnet-5	—	cli	100.0%	100.0%	+0.0 pp	85.7% (6/7)	85.7% (6/7)	1300 → 1300	0.5100 → 0.5100	3720 → 3720	2.1.5 → 2.1.5
fixture-db	codex	gpt-5.6-terra	—	api	85.7%	85.7%	+0.0 pp	71.4% (5/7)	71.4% (5/7)	1200 → 1200	— → —	3600 → 3600	0.121.0 → 0.121.0
fixture-db	codex	gpt-5.6-terra	—	cli	71.4%	71.4%	+0.0 pp	57.1% (4/7)	57.1% (4/7)	1500 → 1500	— → —	4100 → 4100	0.121.0 → 0.121.0

Operational metrics are context only and never affect the verdict or ranking.