docs(results): codify eval result identity contract

[christso]

Summary

• add ADR 0009 for the default experiment bucket and eval_path result identity contract
• update CONCEPTS.md with result source identity and result_dir terminology
• update running-evals docs with experiment precedence and manifest-driven result paths
• mark ADR 0006 as superseded for result bucket/path identity by ADR 0009

Dependency

This PR documents the final result_dir field name and should land after the av-504.2 result_dir rename branch/PR. No code changes are included.

Verification

• bun install
• bun --filter @agentv/web build
• git diff --check
• git diff --cached --check
• rg -n "artifact_dir" CONCEPTS.md apps/web/src/content/docs/docs/evaluation/running-evals.mdx docs/adr/0009-eval-path-result-identity-and-default-experiment.md (no matches)

Biome note: bunx biome check <changed md/mdx files> reports no files processed because Markdown/MDX are outside the configured Biome scope.

[cloudflare-workers-and-pages]

Deploying agentv with    Cloudflare Pages

Latest commit:	da48f18
Status:	✅  Deploy successful!
Preview URL:	https://56cdeb06.agentv.pages.dev
Branch Preview URL:	https://docs-contract-after-result-d.agentv.pages.dev

View logs

[christso]

Findings:

• P2 docs terminology drift: apps/web/src/content/docs/docs/evaluation/running-evals.mdx:113 still describes the generated task bundle as living inside a "per-test artifact directory", and apps/web/src/content/docs/docs/evaluation/running-evals.mdx:363 says the wizard remembers the last run's "artifact directory". This PR is codifying result_dir as the opaque per-row allocation and the run directory as the resume/output boundary, while avoiding artifact_dir as the public term. These two public-doc phrases keep teaching workers to use "artifact directory" for exactly the directories the new contract is trying to name as result/run directories. Suggested fix: change line 113 to "per-test result directory" or "result directory", and line 363 to "last run directory".

Verification performed:

• Fetched current main and PR head with the provided fallback refs (origin/main, refs/remotes/pr/1539), without rewriting remotes.
• Reviewed diff against origin/main for CONCEPTS.md, running-evals.mdx, ADR 0006, and ADR 0009.
• Checked against AGENTS.md, STRATEGY.md, ROADMAP.md, .agents/product-boundary.md, .agents/workflow.md, and .agents/verification.md.
• Ran git diff --check origin/main..refs/remotes/pr/1539.
• Ran git grep -n "artifact_dir" refs/remotes/pr/1539 -- CONCEPTS.md apps/web/src/content/docs/docs/evaluation/running-evals.mdx docs/adr/0009-eval-path-result-identity-and-default-experiment.md docs/adr/0006-separate-experiments-from-eval-definitions.md and found no exact artifact_dir matches in the changed docs.
• Checked GitHub Actions for PR #1539: Build, Typecheck, Lint, Test, Check Links, Validate Evals, Validate Marketplace, and Cloudflare Pages are passing.

Verdict: Ready with a small docs terminology fix; not merging while the finding is unresolved.

[christso]

Unblocked: PR #1540 / av-504.2 has merged into main as c053a9e after live-provider + real LLM-grader dogfood passed.

Please refresh/rebase this branch against current main, rerun checks if GitHub does not do so automatically, and move out of draft when review/verification is complete.