feat: first-party file-source runtime config and provider-agnostic AI settings

[IniZio]

What this PR introduces

• add first-party CLI run commands:
  • skytest run test-case <display-id> --project-id <id> ...
  • skytest run project <project-id> [--display-id ...] ...
• remove bespoke direct run script path and route execution through dispatch/status APIs
• added support for file-source runtime config
• add provider-agnostic team AI runtime config (OpenRouter / OpenAI-compatible)
• harden auth proxy URL/header validation and add local Authgear bootstrap + seed defaults for self-host setups

Step-by-step reviewer guide (from zero)

1. Clone skytest-agent and checkout this branch.
2. Start SkyTest control plane:
   • make bootstrap
   • make dev
3. Keep this running and verify health:
   • curl -fsS http://127.0.0.1:3000/api/health/live
4. Create example project
5. In SkyTest web (http://127.0.0.1:3000), create/get:
   • a project ID for Hanlun (SKYTEST_PROJECT_ID)
   • an API key (SKYTEST_API_KEY)
6. In Hanlun repo, export envs and run:
   • export SKYTEST_BASE_URL=http://127.0.0.1:3000
   • export SKYTEST_PROJECT_ID=<your-project-id>
   • export SKYTEST_API_KEY=<your-api-key>
   • make -C .skytest setup
   • make -C .skytest sync
   • make -C .skytest run-project
7. Expected final result: JSON summary with "status": "PASS" for all selected cases.

Verification in this PR

• npx vitest --run apps/cli/src/cli-parser.test.ts apps/cli/src/commands/init.test.ts
• npm run --workspace @skytest/web test -- src/app/api/test-cases/[id]/route.test.ts src/app/api/test-runs/dispatch/route.test.ts src/lib/runtime/__tests__/team-ai-config.test.ts
• npm run verify (fails only at audit due existing Vite advisories: GHSA-4w7w-66w2-5vf9, GHSA-v2wj-q39q-566r, GHSA-p9ff-h696-f583)