Add cached prompt costs feature with 50% discount

[devin-ai-integration]

Cached Prompt Costs Feature

Implements cached prompt costs with 50% discount as discussed in #215.

Changes

• Added cached_cost_of_token_prompt field to ModelSpec
• Updated register_model! to support cached costs (defaults to 50% of normal prompt cost)
• Modified call_cost to handle cached prompts via new is_cached parameter
• Added comprehensive documentation and examples

Test Plan

1. Create a model with normal and cached costs
2. Verify cost calculation for non-cached prompts
3. Verify cost calculation for cached prompts (50% discount)
4. Test default values and inheritance in model registration

Example:

# Register model with costs
register_model!(
    name = "test-model",
    cost_of_token_prompt = 0.02,     # $0.02 per token
    cost_of_token_generation = 0.03   # $0.03 per token
)

# Test normal vs cached costs
normal_cost = call_cost(100, 50, "test-model")           # Expected: 100 * 0.02 + 50 * 0.03 = 3.5
cached_cost = call_cost(100, 50, "test-model", is_cached=true)  # Expected: 100 * 0.01 + 50 * 0.03 = 2.5

Notes

• Cached cost defaults to 50% of normal prompt cost if not specified
• Backward compatible with existing model registrations
• Documentation updated with examples showing both normal and cached cost calculations

Link to Devin run: https://preview.devin.ai/devin/e70735511ecd43f0a605f7766deecf86

If you have any feedback, you can leave comments in the PR and I'll address them in the app!