feat!: introduce McpObserver for telemetry and message monitoring

[hashemix]

📌 Summary

This PR introduces a new McpObserver trait to the SDK, providing a hook mechanism to intercept incoming and outgoing MCP messages. This allows developers to easily apply telemetry, structured logging, debugging, or custom monitoring solutions across their servers and clients without modifying core business logic.

🔍 Related Issues

• Need request id to generate opentelemetry for mcp server span #133

✨ Changes Made

• Introduced the McpObserver trait with on_receive and on_send hooks.
• Integrated message_observer configuration directly into McpClientOptions and HyperServerOptions (ServerOptions).
• Updated internal client and server runtimes to trigger observer logic on message dispatch and receipt.
• Added comprehensive examples showing how to build remote HTTP loggers: client_observer.rs and server_observer.rs.
• Updated hello-world-mcp-server-stdio and simple-mcp-client-streamable-http examples to utilize the new observers.
• Added a new Message Observer (Telemetry & Monitoring) section to the base and crates README files.
• Added end-to-end tests for observer logic.

🛠️ Testing Steps

1. Run the new observer tests locally: cargo test --package rust-mcp-sdk --test test_observer
2. Launch the hello-world-mcp-server-stdio example and monitor logs using https://app.beeceptor.com/console/rustmcp to see outgoing/incoming server telemetry.

💡 Additional Notes

Operations within on_receive and on_send of the McpObserver trair are on the critical message path. So implementations should either be very fast or spawn an asynchronous, non-blocking task for heavy operations (like HTTP requests) to prevent backpressure.