feat(adk): impl deepagents

[Wenrh2004]

What type of PR is this?

feat(adk): deepagents prebuilt agent with planning, storage, and long-term memory

Check the PR title.

• [x] This PR title match the format: (optional scope):
• [x] The description of this PR title is user-oriented and clear enough for others to understand.
• [x] Attach the PR updating the user documentation if the current PR requires user awareness at the usage level. User docs repo

(Optional) Translate the PR title into Chinese.

feat(adk): 实现 DeepAgents 预置代理，提供规划、存储与长期记忆能力

(Optional) More detailed description for this PR (en / zh).

en: This PR introduces a new prebuilt ADK agent named "DeepAgents" that provides:

• Planning & task decomposition via write_todos (session-scoped plan using planexecute.Plan)
• Storage management tools: ls, read_file, write_file, edit_file (backed by abstract Storage interface for pluggable backends)
• Sub-agent creation via task tool (compose a specialized child agent with custom instruction)
• Long-term memory tools when MemoryStore is provided: remember, recall, update_memory, forget, list_memories (checkpoint-persisted, implementation expected in eino-ext)

Design highlights:

• Agent constructor deepagents.New(ctx, *Config) wires model, storage, optional memory store and extra tools.
• Clear separation between working memory (Session) and long-term memory (MemoryStore).
• Extensible via ToolsConfig to add user-defined tools.
• Secure & predictable tool parameter schemas using schema.ToolInfo; JSON parsing via sonic.

Usage:

• Provide model.ToolCallingChatModel, Storage implementation (eino-ext), and optional MemoryStore.
• Call deepagents.New to obtain an adk.Agent, usable standalone or in compose.StateGraph.

zh: 本 PR 新增预置 ADK 代理 “DeepAgents”，能力包括：

• 规划与任务拆解：write_todos（基于 Session 的 planexecute.Plan）
• 存储管理：ls、read_file、write_file、edit_file（抽象 Storage 接口，便于接入文件系统/内存/S3 等）
• 子代理创建：task 工具（基于给定指令与模型创建子代理）
• 长期记忆（可选）：在提供 MemoryStore 时启用 remember、recall、update_memory、forget、list_memories（基于 CheckPoint 持久化，具体实现建议在 eino-ext）

设计要点：

• 构造函数 deepagents.New(ctx, *Config) 统一注入模型、存储、（可选）记忆库与自定义工具。
• 明确区分工作记忆（Session）与长期记忆（MemoryStore）。
• 通过 ToolsConfig 扩展用户自定义工具。
• 使用 schema.ToolInfo 定义参数协议，sonic 做高性能 JSON 解析。

使用方式：

• 传入 model.ToolCallingChatModel、Storage 实现（eino-ext），以及可选的 MemoryStore。
• 调用 deepagents.New 获得 adk.Agent，可独立使用或作为 StateGraph 节点。

(Optional) Which issue(s) this PR fixes:

(optional) The PR that updates user documentation: