feat(results): export remote runs and sync Studio

[christso]

Closes #826
Closes #992

Summary

• add results.export config support and shared cached git repo utilities for draft-PR artifact export
• auto-export eval and pipeline bench artifacts to a configured remote repo with graceful warnings on auth/access failures
• extend Studio backend and UI to merge local/remote runs, tag source, expose sync/status endpoints, and support source filtering with local-only fallback

Verification

• bun test packages/core/test/evaluation/loaders/config-loader.test.ts packages/core/test/evaluation/validation/config-validator.test.ts apps/cli/test/commands/results/serve.test.ts
• bunx tsc -p apps/studio/tsconfig.json --noEmit
• git diff --check

Notes

• bunx tsc -p apps/cli/tsconfig.json --noEmit is currently noisy due existing @agentv/core declaration-resolution issues in the repo, so it was not a useful signal for this change.

[cloudflare-workers-and-pages]

Deploying agentv with    Cloudflare Pages

Latest commit:	ce01810
Status:	✅  Deploy successful!
Preview URL:	https://06352432.agentv.pages.dev
Branch Preview URL:	https://feat-826-992-results-export.agentv.pages.dev

View logs

[christso]

E2E Verification Results

All pre-push hooks pass (build, typecheck, lint, test, validate) after the DEFAULT_THRESHOLD fix.

1. Eval run — no regression, no-op when unconfigured

$ bun apps/cli/src/cli.ts eval examples/features/basic/evals/dataset.eval.yaml --dry-run --test-id code-review-javascript
Artifact directory: .agentv/results/runs/default/2026-04-09T00-16-30-674Z
Results written to: .agentv/results/runs/default/2026-04-09T00-16-30-674Z/index.jsonl

Run completes and writes artifacts; no remote-export errors (no config → silent no-op). ✅

2. Studio — local runs have source: 'local'

{"runs":[{"filename":"2026-04-09T00-16-30-674Z","source":"local","target":"llm-dry-run",...}]}

✅

3. Remote status — sensible response when unconfigured

{"configured":false,"available":false,"repo":"","cache_dir":"","run_count":0}

✅

4. Results export command

Exported 1 test(s) to /tmp/export
  code-review-javascript

✅

5. All 1953 tests pass (core: 1461, eval: 67, cli: 425) ✅

---

Code Review Findings & Fixes

One fix applied before marking ready:

Fixed: pipeline/bench.ts was hardcoding 0.8 for pass/fail in the export summary instead of importing and using DEFAULT_THRESHOLD. Fixed in commit 7221f6a9.

Not a bug: pipeline/run.ts intentionally does not auto-export — it's a pre-grading step that writes response.md files, not final index.jsonl results. Auto-export correctly happens at pipeline bench which produces the final JSONL.