Add global rate limiting middleware with Retry-After headers

[artugro]

Problem

The current quota system only limits per-integration invocations via the broker. There is no global rate limiting at the HTTP layer, meaning:

• No protection against burst traffic from any endpoint
• No Retry-After headers to guide clients on backoff
• No per-IP or per-user request rate limiting
• Health check and public endpoints have no protection

Proposed Solution

Add a Redis-backed sliding window rate limiter as FastAPI middleware:

• Use Redis sorted sets or token bucket algorithm
• Return 429 Too Many Requests with Retry-After header
• Configurable per-route or global limits via settings
• Graceful degradation when Redis is unavailable (allow all)
• Consider using slowapi library or building on existing Redis infrastructure

Add settings: RATE_LIMIT_ENABLED, RATE_LIMIT_REQUESTS_PER_MINUTE, RATE_LIMIT_BURST.

Files

• New: src/core/rate_limit.py
• src/main.py (add middleware)
• src/core/settings.py (add settings)