fix(cost-model): cap Value generators to prevent unrealistic benchmark sizes

Refactor Value benchmark generators to use Cardano-compliant constraints
and prevent generation of unrealistically large Values that were causing
inflated cost model parameters.

- Fix policy IDs to exactly 28 bytes (MintingPolicyHash size)
- Limit token names to 0-32 bytes per Cardano ledger spec
- Cap random generation to max 10,000 policy/token pairs
- Simplify generator architecture by removing complex layered logic
- Remove unused helper functions and imports
- Apply code quality improvements from HLint suggestions

This addresses the unValueData builtin having an extremely high intercept
cost (10+ billion) due to benchmarking against Values with millions of
entries that would never exist on-chain.

Author: Unisay