Feature Request: Integrate Anthropic's New Prompt Caching

[stchakwdev]

We should consider incorporating Anthropic's new prompt caching feature into this project. This integration could offer several benefits:

1. Reduced costs: By caching prompts, we can potentially lower the overall API usage costs.
2. Improved latency: Cached prompts should result in faster response times for repeated or similar queries.
3. Enhanced "Agentic" capabilities: Anthropic claims this feature improves their model's ability to act as an agent. It would be interesting to explore how this affects Devon's performance and capabilities.

Implementing this feature could significantly enhance our project's efficiency and potentially expand its functionality.