Add notebook: Building Low-Latency AI Agent Workflows with Amazon Bedrock Prompt Caching

[esbjj]

This PR adds a notebook demonstrating how to optimize AI agent workflows using Amazon Bedrock's prompt caching capabilities for production deployments.

What this notebook covers:

• Implementation of efficient AI agent workflows with prompt caching
• Performance optimization techniques achieving up to 85% latency reduction and 90% cost savings
• Identification and caching of static prompt components (system instructions, tool definitions)
• Performance monitoring and analysis utilities
• Integration patterns with Claude 3.7 Sonnet model

Key benefits demonstrated:

• Reduced token consumption through caching of static prompt components
• Significant latency improvements for agent interactions
• Cost optimization for production-scale deployments
• Improved throughput for concurrent users

Prerequisites:

• AWS account with Amazon Bedrock access
• Access to Anthropic Claude 3.7 Sonnet model
• Python 3.7+
• Basic understanding of LLMs and prompt engineering

The notebook is designed for sequential execution and includes practical examples comparing cached vs non-cached implementations with performance metrics.