This PR improves and cleans up the benchmarking code in the following ways:

• Removed outdated JSON benchmarks that provided no value or useful comparisons. A proper comparative benchmark can be added later
• Organized all benchmarks into a single group without separate benchmarking functions
• Proper usage of parameterization and black boxes for cleaner code structure
• Replaced deprecated criterion black box with the std one
• Added throughput measurement for additional metrics
• Removed the deprecated paste macro

The benchmarks themselves remain unchanged. Only the code organization has been improved for better maintainability and consistency.