[Feature Request] Native support for Custom / Client-Executed Tool Search with Prompt Caching and defer_loading awareness in tool selection middlewares

### Submission checklist - [x] This is a feature request, not a bug report or usage question. - [x] I added a clear and descriptive title that summarizes the feature request. - [x] I used the GitHub search to find a similar feature request and didn't find it. - [x] I checked the LangChain documentation and API reference to see if this feature already exists. - [x] This is not related to the langchain-community package. ### Package (Required) - [x] langchain - [x] langchain-openai - [x] langchain-anthropic - [ ] langchain-classic - [ ] langchain-core - [ ] langchain-model-profiles - [ ] langchain-tests - [ ] langchain-text-splitters - [ ] langchain-chroma - [ ] langchain-deepseek - [ ] langchain-exa - [ ] langchain-fireworks - [ ] langchain-groq - [ ] langchain-huggingface - [ ] langchain-mistralai - [ ] langchain-nomic - [ ] langchain-ollama - [ ] langchain-openrouter - [ ] langchain-perplexity - [ ] langchain-qdrant - [ ] langchain-xai - [ ] Other / not sure / general ### Feature Description Native support for Custom / Client-Executed Tool Search with Prompt Caching and defer_loading awareness within LangChain's tool selection middlewares (LLMToolSelectorMiddleware and ProviderToolSearchMiddleware). ### Use Case When building agents with large toolsets (30+ tools), passing all tool schemas on every model call introduces significant token bloat and degrades latency. More importantly, it breaks Prompt Caching for providers like OpenAI and Anthropic, because any dynamic variation in the tool definitions invalidates the cached prefix. According to official provider guidelines: [Anthropic Custom Tool Search Implementation](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool#custom-tool-search-implementation) [OpenAI Client-Executed Tool Search](https://developers.openai.com/api/docs/guides/tools-tool-search#client-executed-tool-search) The recommended architecture separates tools into Always-Included Core Tools (static prefix to maximize cache…