Rate limits clarified

The documentation now specifies both the per-token and per-IP limits and explains how shared jobs should handle them.

The rate-limit guide now states both limits:

• 500 requests per token per minute.
• 500 requests per IP per minute.

Both apply to the same traffic. The guide covers coordinating shared jobs, reusing snapshots and handling 429 responses. This corrects the documented limits; it is not an announcement of a new API limit.