feat(provider): add CodeBuddy international provider (www.codebuddy.ai)

[Yanu403]

Summary

Adds CodeBuddy (www.codebuddy.ai) as a fully functional provider with API key authentication, enabling users to access GPT-5.5, GPT-5.4, Claude Sonnet 4.6, and Gemini 3.1 Pro through their CodeBuddy subscriptions.

The existing codebuddy.js registry entry was hidden: true and pointed to copilot.tencent.com (China-only). This PR switches it to the international domain, adds models, and exposes it to users.

Changes

1. Registry (providers/registry/codebuddy.js)

• Switch baseUrl from copilot.tencent.com to www.codebuddy.ai/v2/chat/completions
• Change category from oauth to apikey (API key is the primary auth method)
• Add forceStream: true (server requires streaming — rejects non-stream with error 11101)
• Add 4 confirmed-working models: gpt-5.5, gpt-5.4, claude-sonnet-4.6, gemini-3.1-pro
• Set hidden: false to expose to users
• Keep OAuth config for future device flow support

2. Executor (executors/codebuddy.js)

Lightweight shim over DefaultExecutor (~53 LOC). Two responsibilities:

• System message injection: CodeBuddy's API requires a system message in the messages array (returns 11101:invalid request without one). Injects You are a helpful assistant. when absent.
• Dual-header auth: Sends both Authorization: Bearer and X-API-Key with the same token, matching the CLI's fingerprint.

3. Executor registration (executors/index.js)

Import and register CodeBuddyExecutor in the executor map.

API Wire Format

Captured via mitmproxy interception of codebuddy -p CLI v2.106.3:

Request:

POST https://www.codebuddy.ai/v2/chat/completions
Authorization: Bearer ck_<id>.<secret>
Content-Type: application/json

{
  "model": "gpt-5.5",
  "messages": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Hello"}
  ],
  "stream": true
}

Response (SSE):

data: {"id":"resp_...","model":"gpt-5.5","object":"response","choices":[{"delta":{"content":"Hi"}}]}

data: {"id":"resp_...","choices":[{"delta":{"content":""},"finish_reason":"stop"}],"usage":{"total_tokens":42}}

data: [DONE]

Testing

• ✅ Syntax validation (Node.js ESM import check)
• ✅ Unit tests: system message injection (4/4 pass — inject, no-duplicate, empty-messages, headers)
• ✅ Live smoke test: real API call through executor → HTTP 200 → reconstructed "PONG"
• ✅ Verified with ck_* API key from active CodeBuddy account

API Key Setup

Users obtain API keys from: https://www.codebuddy.ai/console/api-keys

Key format: ck_<id>.<secret>

Add as API key connection in 9router dashboard → Providers → CodeBuddy.

Constraints (documented in executor JSDoc)

1. System message required: Server returns {"code":11101,"msg":"Parse message failed: 11101:invalid request"} without it. Handled by ensureSystemMessage() in the executor.
2. Stream-only: stream: false returns {"code":11101,"msg":"Non-stream chat request is currently not supported"}. Handled by forceStream: true in registry.

Not in scope (future PRs)

• Device flow OAuth (/v2/plugin/auth/state + /v2/plugin/auth/token)
• Live model discovery via dashboard API
• Tool calling validation (presumably zero-effort since OpenAI format, not yet tested)
• Vision/image input support

---

Closes #