Add token cost optimization cookbook with 5 practical techniques

[Ali-mohamed843]

Description

This cookbook teaches developers 5 practical techniques to reduce
Claude API token costs with real before/after measurements and dollar
calculations.

Closes #704 (or whatever your issue number is)

What this cookbook demonstrates

• Prompt Compression — remove filler and measure token savings
• Conversation Summarization — condense long chat histories to
  reduce context size
• Intelligent Model Routing — classify task complexity and route
  to cheapest adequate model
• Prompt Caching — use cache_control for 90% off repeated system
  prompts
• Batch API — 50% discount for non-urgent workloads

Each technique includes working code, before/after token counts, and
exact dollar cost comparisons. Section 7 combines all 5 techniques
and shows the cumulative impact with a bar chart.

Type of Change

• New cookbook

Checklist

• Notebook runs end to end with a valid ANTHROPIC_API_KEY
• Offline fallback included for users without API access
• No hardcoded API keys
• Placed in costs_and_performance/ folder
• Follows existing notebook style

[Ali-mohamed843]

Note: Sections on prompt caching and batch API reference existing cookbooks in misc/ and can be removed if the maintainers prefer. The core original contributions are the prompt compression, conversation summarization, and model routing techniques combined with the cumulative cost impact analysis in Section 7.