feat: 新增独立 browser 命令组以支持 CDP 启动与双后端运行

[duo121]

Summary

Add an additive opencli browser command group for browser backend selection, CDP browser launch, and command forwarding, without modifying existing command surfaces.

Why

From an AI-native CLI perspective, browser access should be a runtime capability that can be selected explicitly, instead of being hard-wired to one integration path.

This matters for two reasons:

1. Existing browser commands can be reused against a Chrome DevTools Protocol endpoint.
2. The change can stay non-invasive: no breaking changes to opencli <site> <command>.

Proposed Commands

• opencli browser launch
  • launch a Chrome/Chromium instance with a CDP debugging port
  • support optional --user-data-dir
  • default to an isolated profile when not provided
• opencli browser doctor
  • diagnose extension or cdp backend connectivity
• opencli browser run -- <existing command>
  • forward an existing opencli command through a selected browser backend

Design Principles

• additive only: do not rewrite existing commands
• preserve the current execution architecture
• make backend choice explicit via runtime flags
• keep the old plugin/extension path working

Example

opencli browser launch --port 9222
opencli browser doctor --backend cdp --cdp-endpoint http://127.0.0.1:9222 --live
opencli browser run --backend cdp --cdp-endpoint http://127.0.0.1:9222 -- zhihu search --keyword AI

Notes

This is intentionally a separate command group so the maintainer can evaluate whether and how these capabilities should be merged into the main command surface later.