Fix busy-spin CPU pin after stdio transport disconnects

[bessavagner]

Summary

• run_server() had a trailing while True: pass meant to "keep the process alive" after mcp.run() returns.
• mcp.run() already blocks until the transport shuts down, so reaching that code means the server has already stopped (e.g. the stdio client disconnected).
• The empty spin loop has no sleep/yield, so it pins one CPU core at 100% indefinitely until the process is manually killed.
• Easy to reproduce: any MCP client that spawns this server over stdio and later disconnects leaves an orphaned process spinning at 100% CPU forever. I hit this myself — the process had been running at 100% CPU for over two days before I noticed.

Fix

Remove the busy-spin loop and let run_server() exit cleanly with a log line once mcp.run() returns, since there's nothing left to do at that point.

Test plan

• Read through run_server() control flow to confirm mcp.run() is blocking and the trailing code only runs after shutdown
• Ran the server locally over stdio and confirmed it now exits cleanly on disconnect instead of spinning