Implement automatic retry logic for LLM rate limits (HTTP 429) to ensure workflow continuity

Here is a suggestion based on my usage experience:

In your development and testing scenarios, token quotas are likely plentiful. However, in actual production use, I've found that when the LLM API hits its rate limit (HTTP 429), the Hiclaw bot halts the entire workflow. The process stops completely, requiring me to manually re-mention the bot to resume work.

I attempted to add custom rules instructing the bot to retry after receiving a 429 error, but its adherence to these instructions has been inconsistent and unreliable.

In contrast, tools like claude-code handle this scenario exceptionally well. Their automatic retry mechanism after a 429 error allows for a truly hands-off experience; I can simply check back later to verify the results without needing to monitor the process.

This issue significantly impacts usability. My primary goal in using this tool is automation, and having workflows break due to rate limits defeats that purpose. Implementing robust, native retry logic would greatly improve the reliability of Hiclaw for automated tasks.