fix: 主线 B3/B4 结构化工具调用与完成前验证

- 打通 Worker → WorkerManager → Provider 的 tools 传递链路，ProviderManager/adapter
  返回结构化 tool_calls 给 WorkerRuntime
- OpenAI-compatible/Anthropic adapter 发送工具 schema，并解析 provider 返回的
  tool_calls/tool_use；OpenAI 工具名使用 fs.write ↔ fs__write 双向映射
- 修复独立复审发现的 OpenAI 协议隐患：assistant tool_use blocks 必须转换为
  assistant.tool_calls，后续 role=tool 消息的 tool_call_id 必须匹配前一轮
  tool_calls[].id；不再把 tool_use JSON 字符串化为普通文本
- ExecutorRole 优先消费原生 tool_calls，回灌 canonical tool_result block；移除
  fs.write(...)/shell.run(...) 函数调用正则解析，只保留严格 JSON tool_call
  fallback 与 filename code block 兼容
- DONE 前执行 verification-before-completion：任务要求 build/compile/run/test/编译/
  运行/测试时必须实际 shell.run 验证，失败不 checkpoint、不返回 completed
- fs.write 覆盖已有文件也强制 read-before-write，补齐 Claude Code 文件状态纪律
- 新增 packages/workers/test/executor-role.test.ts 行为测试：原生 tool_calls 执行、
  verification 失败不得 completed

真实验收:
- TSC=0
- bun test packages/workers/test/executor-role.test.ts: 2 pass / 0 fail
- OpenAI converter 探针确认 assistant.tool_calls 与 role=tool 的 tool_call_id 匹配
- 真实 GLM Worker C++ 编译运行任务通过，worker verification 记录实际命令:
  c++ hello.cpp -o /tmp/aircoding-verify && /tmp/aircoding-verify

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

packages/llm/src/adapters/AnthropicAdapter.ts

@@ -41,6 +41,7 @@ interface AnthropicApiRequest {
   top_p?: number
   system?: string
   stream?: boolean
+  tools?: Array<{ name: string; description: string; input_schema: Record<string, unknown> }>
 }
 
 export class AnthropicAdapter implements ProviderAdapter {
@@ -101,6 +102,7 @@ export class AnthropicAdapter implements ProviderAdapter {
       temperature: input.temperature,
       system: input.system as string | undefined,
       stream: false,
+      tools: this.convert_tools(input.tools),
     })
 
     // Yield each content block as an event
@@ -110,6 +112,8 @@ export class AnthropicAdapter implements ProviderAdapter {
         yield { type: 'content_delta', payload: { type: 'text_delta', text: block.text } }
       } else if (block.type === 'thinking' && block.thinking) {
         yield { type: 'content_delta', payload: { type: 'thinking_delta', thinking: block.thinking } }
+      } else if (block.type === 'tool_use' && block.name) {
+        yield { type: 'tool_use', payload: { id: block.id, name: block.name, input: (block.input as Record<string, unknown>) || {} } }
       }
     }
     if (response.usage) {
@@ -126,19 +130,36 @@ export class AnthropicAdapter implements ProviderAdapter {
    * Backward-compat: single-shot complete that returns string content.
    * Used by MainAgent.classify_via_llm.
    */
-  async complete_text(messages: unknown[], options: { model?: string; max_tokens?: number } = {}): Promise<{ content: string; usage?: { input_tokens: number; output_tokens: number } }> {
+  async complete_text(messages: unknown[], options: { model?: string; max_tokens?: number; tools?: unknown[] } = {}): Promise<{ content: string; usage?: { input_tokens: number; output_tokens: number }; tool_calls?: Array<{ id?: string; name: string; arguments: Record<string, unknown> }> }> {
     const response = await this.make_request({
       model: options.model || 'claude-haiku-4-5-20251001',
       messages: this.convert_raw_messages(messages),
       max_tokens: options.max_tokens || 1024,
+      tools: this.convert_tools(options.tools),
       stream: false,
     })
+    const tool_calls = response.content
+      .filter(b => b.type === 'tool_use' && b.name)
+      .map(b => ({ id: b.id, name: b.name!, arguments: (b.input as Record<string, unknown>) || {} }))
     return {
       content: this.extract_content(response),
       usage: response.usage ? { input_tokens: response.usage.input_tokens, output_tokens: response.usage.output_tokens } : undefined,
+      tool_calls: tool_calls.length ? tool_calls : undefined,
     }
   }
 
+  private convert_tools(tools?: unknown[]): Array<{ name: string; description: string; input_schema: Record<string, unknown> }> | undefined {
+    if (!tools?.length) return undefined
+    return tools.map(t => {
+      const tool = t as { name: string; description?: string; input_schema?: Record<string, unknown> }
+      return {
+        name: tool.name,
+        description: tool.description || tool.name,
+        input_schema: tool.input_schema || { type: 'object', properties: {} },
+      }
+    })
+  }
+
   private convert_to_anthropic_messages(messages: Array<{ role: string; content: unknown }>): Array<{ role: string; content: Array<Record<string, unknown>> }> {
     return messages.map(m => {
       const blocks: Array<Record<string, unknown>> = []