feat: Add tool annotations for improved LLM tool understanding

[bryankthompson]

Summary

Adds MCP tool annotations (readOnlyHint, title) to all 5 recipe query tools to help LLMs understand that these are safe, read-only operations.

Changes

All tools receive readOnlyHint: true annotation with English titles:

Tool	Title	Annotation
mcp_howtocook_getAllRecipes	Get All Recipes	readOnlyHint: true
mcp_howtocook_getRecipesByCategory	Get Recipes by Category	readOnlyHint: true
mcp_howtocook_getRecipeById	Get Recipe by ID	readOnlyHint: true
mcp_howtocook_recommendMeals	Recommend Weekly Meals	readOnlyHint: true
mcp_howtocook_whatToEat	What to Eat	readOnlyHint: true

Why This Matters

Tool annotations provide semantic metadata that:

• Enables MCP clients to auto-approve safe read-only operations
• Improves tool selection accuracy
• Required for MCP Directory compliance

Testing

• Build passes (npm run build)
• tools/list returns all 5 tools with annotations
• All annotations match actual tool behavior (all are read-only queries)

Before/After

Before:

server.tool(
  "mcp_howtocook_getAllRecipes",
  "获取所有菜谱",
  { /* schema */ },
  async () => { /* handler */ }
);

After:

server.tool(
  "mcp_howtocook_getAllRecipes",
  "获取所有菜谱",
  { /* schema */ },
  {
    title: "Get All Recipes",
    readOnlyHint: true,
  },
  async () => { /* handler */ }
);