feat(cost-model): update benchmark data for Value builtin functions

Regenerate benchmark measurements for lookupCoin, valueContains,
valueData, and unValueData builtins. The updated benchmarks reflect
optimized test cases with realistic Value sizes and proper parameter
coverage for Conway-era cost modeling.

The new benchmark data shows significantly improved performance
characteristics for Value operations, particularly for valueContains
and unValueData which previously had placeholder costs.

Author: Unisay