[integer] Optimize BigUInt addition and subtraction with SIMD and early stop tricks

[forfudan]

This pull request introduces significant updates to the bench_biguint benchmarking suite, optimizes arithmetic operations in the BigUInt and BigDecimal modules, and refactors method names for clarity and consistency. Additionally, new benchmarking cases and constants are added to improve performance testing and support for larger numbers.

1. Use SIMD to accelerate BigUInt addition and in-place addition. The speed gain is 2x to 4x for large numbers.
2. Refine the BigUInt subtraction and in-place addition with some tricks on carry so that floor_divide and modulo are replaced by addition and subtraction.
3. Use a trick to first do a paralelled addition word-by-word, and then do normalized carries with one loop.

Arithmetic Optimizations:

• src/decimojo/bigdecimal/arithmetics.mojo: Refactored methods to replace scale_up_by_power_of_10 and scale_down_by_power_of_10 with multiply_by_power_of_ten and floor_divide_by_power_of_ten, improving naming consistency and clarity. [1] [2] [3]
• src/decimojo/biguint/biguint.mojo: Removed redundant add_inplace_by_1 method and replaced it with a more general add_inplace_by_uint32 for optimized addition operations. [1] [2]

Refactoring and Enhancements:

• src/decimojo/biguint/biguint.mojo: Renamed methods (e.g., scale_up_by_power_of_10 → multiply_by_power_of_ten) for consistency across the codebase and introduced VECTOR_WIDTH constant for SIMD-based arithmetic optimizations. [1] [2]
• src/decimojo/bigdecimal/comparison.mojo: Updated comparison logic to use the newly renamed multiply_by_power_of_ten method for scaling coefficients.

Benchmarking Updates:

• benches/biguint/bench_biguint_add.mojo: Added five new addition benchmark cases for larger word sizes (e.g., 4096 words + 2048 words) to test scalability.
• benches/biguint/bench_biguint_multiply.mojo: Expanded multiplication benchmarks to include 12 new cases with varying word sizes, introducing reduced iterations for very large numbers to optimize runtime.
• benches/biguint/bench_biguint_multiply_complexity.mojo: Adjusted test sizes to start from 8 words instead of 32 and updated iteration logic for benchmarking complexity. [1] [2]

These changes collectively enhance the code's readability, scalability, and performance, especially for operations involving large numbers and benchmarking scenarios.