Feature request: native Rate Limiting (MCP spec compliance)

Problem

The MCP specification requires rate limiting across multiple message types.

Required (MUST/SHOULD):

• Progress notifications: "Both parties SHOULD implement rate limiting to prevent flooding"
• Completion requests: "MUST implement appropriate rate limiting"
• Sampling requests: "Clients SHOULD implement rate limiting"
• Logging messages: "Servers SHOULD rate limit log messages"
• Elicitation requests: "Clients SHOULD implement rate limiting"

Proposed Architecture

Transport Layer Rate Limiting (token bucket algorithm)

Why Transport Layer:

• Universal Coverage: All message types, all transports (stdio/SSE/HTTP/WebSocket)
• Single Integration Point: All messages flow through Transport::send()
• Existing Patterns: Following established decorator pattern (SinkStreamTransport)
• Performance: Early filtering, concurrent-safe

Start at the transport layer and trace how messages pass through Transport::send(), then read the existing SinkStreamTransport decorator pattern. Define how a concurrent-safe token bucket applies to Progress, Completion, Sampling, Logging, and Elicitation messages across stdio, SSE, HTTP, and WebSocket transports. Done means all listed message types receive the required transport-level rate limiting.