docs(shopping): add MCP method, param, and result descriptions for agent tooling parity

[prasad-stripe]

Summary

The MCP (JSON-RPC) service definition for Shopping lacked descriptions on most methods, all params, and all results. The REST (OpenAPI) counterpart already has these. This creates a parity gap that directly impacts AI agent usability: when an LLM reads the MCP tool definitions, descriptions are the primary signal for understanding what a method does, what each parameter means, and what to expect in the result.

This PR adds:

• Method-level descriptions to the 7 methods that were missing them (mirroring REST)
• Param descriptions to all 30 parameters across all 13 methods
• Result descriptions to all 13 methods
• request-id (UUID) to the meta schema, mapping to the HTTP Request-Id header for cross-transport tracing

Why this matters

MCP is the agent-native transport. AI agents consuming these tools via JSON-RPC rely on description fields to determine which method to call, what parameters to provide, and how to interpret results. Without descriptions, agents must infer behavior from names alone, leading to misuse, unnecessary retries, and degraded user experiences.

The REST spec already has this information; the MCP spec should not be a second-class citizen for the same service.

Test plan

• Verify valid OpenRPC 1.3.2 JSON (no schema validation CI currently, but structure is correct)
• Confirm descriptions match REST counterparts semantically
• No behavioral changes; documentation only