Support for seperate retry logic for rate limit errors and structured output validation errors

[tannonk]

Problem

Gemini's free plan offers rather restrictive rate limits (e.g. 15 RPM). However, the retry logic appears to support validation errors OR API Errors.

Solution
Ideally, it would be great to have an example which shows how to best separate these such that they can be handles differently (e.g. exponential backoff for rate-limit-related errors, faster retries for validation errors).