fix: preserve native ripgrep capabilities for direct models

Replace the restricted workspace_rg parameter mapping with native args and optional cwd. Preserve rg output and exit codes, and reuse process cancellation and output_read pagination so large search results remain accessible. Ask remains read-only within the workspace; Execute uses current-account permissions.

Validation: focused regression 169 passed, 1 skipped; typecheck and lint passed. Real DeepSeek Ask and Execute pagination scenarios passed with six counted requests in the successful run; an earlier failed fixture run has no retained request count. Full-suite validation is not green: the recorded run had unrelated failures and the latest run timed out without a summary. The remote Agent does not use this local execution path.

Release note: 直连模型可使用 ripgrep 原生参数搜索文件并保留原生结果，大输出可通过 output_read 完整续读。Ask 保持工作区只读，Execute 使用当前账号权限。

src/main/agent/model-runtime.test.ts

@@ -2234,7 +2234,7 @@ describe('ModelAgentRuntime', () => {
       const reference = previous.stdoutReference!
       expect(reference).toBeDefined()
       const calls = [
-        { name: 'workspace_rg', arguments: { pattern: 'readonly-marker', glob: ['README.md'] } },
+        { name: 'workspace_rg', arguments: { args: ['-g', 'README.md', 'readonly-marker'] } },
         { name: 'workspace_read_text', arguments: { path: 'README.md' } },
         { name: 'output_read', arguments: { handle: reference.handle, cursor: reference.nextCursor } }
       ]
@@ -2261,7 +2261,7 @@ describe('ModelAgentRuntime', () => {
         (message: { role: string }) => message.role === 'tool'
       )
       expect(results).toHaveLength(3)
-      expect(results[0].content).toContain('README.md:1:1:readonly-marker')
+      expect(results[0].content).toContain('README.md:1:readonly-marker')
       expect(results[1].content).toContain('readonly-marker')
       expect(JSON.parse(results[2].content)).toMatchObject({
         content: expect.stringContaining('retained-tail'), eof: true