fix(providers): support Copilot SDK OpenAI endpoint parity runs

[christso]

Summary

• update @github/copilot-sdk to v1.0.3 and adapt AgentV to the v1 session/client lifecycle
• add Copilot SDK BYOK model_id/wire_model support for local OpenAI-compatible endpoints
• move workspace after_each reset until after graders inspect the agent-modified workspace
• keep separate Codex SDK local-endpoint support/config tests

Verification

• bun test packages/core/test/evaluation/providers/copilot-sdk.test.ts packages/core/test/evaluation/providers/targets.test.ts packages/core/test/evaluation/validation/targets-validator.test.ts
• bun --filter @agentv/core typecheck
• bun --filter agentv typecheck
• bun --filter @agentv/core build && bun --filter agentv build
• From next-evals-oss-agentv parity harness: npm run agentv:validate; npm run agentv:dry; timeout 240s npm run agentv:run; npm run compare-results -- --vercel vercel/results --agentv .agentv/results/runs/agentv-parity --agentv-experiment agentv-copilot-sdk-openai --export tmp/real-agent-results.json

Live parity result

The AgentV Copilot SDK smoke passed the tiny Next.js fixture at 100% using the local OpenAI-compatible endpoint with model/model_id=gpt-5 and wire_model=gpt-5.3-codex-spark. Comparison against existing Vercel results passed.

[cloudflare-workers-and-pages]

Deploying agentv with    Cloudflare Pages

Latest commit:	5bd477b
Status:	✅  Deploy successful!
Preview URL:	https://944bb0d3.agentv.pages.dev
Branch Preview URL:	https://fix-openai-codex-target-env.agentv.pages.dev

View logs