Reevaluate the use of cache in the gas test scripts

We've encountered issues where values cached across different test runs cause unexpected test failures due to transactions being invalid (e.g. trying to burn a wrong token_id that doesn't exist but was saved from previous runs).

Separately, there are some items that are saved to cache but are never retrieved (like `evm::deterministic::contract`, which only add complexity when reading the code.

We should reevaluate whether the cache is necessary and needed for our intended use case - running the script periodically on CI.

Expected outcome is to either:
- remove the cache, if its use is unjustified
- explain (in a comment) why the cache is worthy, and how it can be improved so that it doesn't cause issues (possibly create a follow up ticket)

Start with the gas test scripts and trace where cached values are written and retrieved, including the evm::deterministic::contract entry. Run the scripts across repeated test runs to reproduce stale transaction data. Done means the cache is removed, or its purpose and safeguards are documented with any follow-up work identified.

Written by the indexing model from the issue text.