Feature Request: Pass `_meta` parameter through to MCP SDK `call_tool()`

### Problem When invoking MCP tools via `langchain-mcp-adapters`, the `_meta` parameter is not passed through to the MCP SDK's `call_tool()` method, even though: 1. The MCP SDK already supports it - `ClientSession.call_tool()` accepts `meta` as a keyword argument 2. The MCP specification (2025-06-18) defines `_meta` as a reserved field for attaching metadata to requests This prevents users from leveraging MCP's metadata capabilities for server-specific context. ### Use Case MCP servers may need request-level metadata for: - **Custom routing** - directing requests to specific backends or tenants - **Session/request context** - passing correlation IDs, user context, etc. - **Server-specific features** - any metadata the server needs per-invocation Currently, `headers` can be passed at the connection level (static), but there's no way to pass dynamic, per-request metadata via `_meta`. ### Proposed Solution 1. Add `meta` field to `MCPToolCallRequest` dataclass 2. Extract `_meta` from tool arguments and pass it to `session.call_tool()` 3. Allow interceptors to modify `meta` via `request.override(meta={...})` Example usage: ```python # Pass _meta as part of tool arguments result = await tool.ainvoke({ "message": "hello", "_meta": {"session_id": "abc123"} }) # Or via interceptor async def add_context_interceptor(request, handler): modified = request.override(meta={"user_id": "current-user"}) return await handler(modified) ``` ### MCP Spec Reference Per the [MCP specification](https://modelcontextprotocol.io/specification/2025-06-18/basic/index#meta): > The `_meta` property/parameter is reserved by MCP to allow clients and servers to attach additional metadata to their interactions. ### Implementation I have a working implementation with tests ready to submit as a PR if this feature is accepted.