feat(pi-web-search): 统一 web_search/web_fetch 并由主模型路由搜索厂商

7 个按厂商命名的搜索工具收敛为 2 个统一工具，schema 与提示词不再暴露
provider 概念：web_search 由当前主模型按内置规则在 Tavily/Exa/Keenable
间路由（会话内 LRU 缓存，失败按可用性降级），web_fetch 固定走 Exa 并
静默降级。三家均改为纯 fetch 直连——Exa 直接对 mcp.exa.ai 发 JSON-RPC
并缓存会话，彻底移除 pi-mcp-adapter、@tavily/pi-extension 与
@keenable/pi-search 依赖。

Generated with [Devin](https://devin.ai)

Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>

pi-tool-search/extensions/bundle-groups.ts

@@ -77,31 +77,13 @@ export const BUNDLE_GROUP_DEFINITIONS: BundleGroupDefinition[] = [
     tools: ["codegraph_explore", "lsp_definition", "lsp_references", "lsp_hover", "lsp_symbols", "lsp_diagnostics"],
   },
   {
-    id: "web-tavily",
-    title: "Tavily web discovery",
-    summary: "Discover broad, current, or news-oriented web sources with Tavily and fetch selected pages.",
-    useWhen: ["Keywords are uncertain, many candidate sources help, or current/news coverage is needed"],
-    avoidWhen: ["Official technical sources or Chinese site/date constraints are more important"],
-    keywords: ["tavily", "web", "news", "discovery", "网页", "新闻", "广泛搜索"],
-    tools: ["tavily_web_search", "tavily_web_fetch"],
-  },
-  {
-    id: "web-exa",
-    title: "Exa precision research",
-    summary: "Find precise official, technical, academic, company, or API sources with Exa and read selected pages.",
-    useWhen: ["Source precision, official documentation, papers, releases, or advanced filters matter"],
-    avoidWhen: ["Broad news discovery or Chinese policy search is the primary need"],
-    keywords: ["exa", "official docs", "paper", "api", "research", "官方文档", "论文", "精确搜索"],
-    tools: ["exa_web_search", "exa_web_search_advanced", "exa_web_fetch"],
-  },
-  {
-    id: "web-keenable",
-    title: "Keenable focused search",
-    summary: "Search Chinese-language, site-restricted, policy, government, or date-filtered sources and fetch selected pages.",
-    useWhen: ["Chinese pages, policies, government notices, domain restrictions, or publication dates matter"],
-    avoidWhen: ["Broad international discovery or academic semantic search is a better fit"],
-    keywords: ["keenable", "Chinese", "policy", "site", "date", "中文", "政策", "站点", "日期"],
-    tools: ["keenable_search", "keenable_fetch"],
+    id: "web-search",
+    title: "Web search and fetch",
+    summary: "Search the web for current or citable information and read selected pages as clean markdown.",
+    useWhen: ["Current events, external facts, documentation, papers, or page content are needed"],
+    avoidWhen: ["The answer is already in the workspace or local code tools suffice"],
+    keywords: ["web", "search", "fetch", "news", "docs", "网页", "搜索", "新闻", "文档"],
+    tools: ["web_search", "web_fetch"],
   },
   {
     id: "context-execution",