bugfix(sierra-sim): Fixed u128_guarantee_mul simulation. - #10050

Summary

Fixes the u128_guarantee_mul simulation to correctly compute the high and low 128-bit limbs of a 128×128-bit multiplication. The division was previously using BigInt::one().pow(128) (which equals 1, not 2^128) instead of BigInt::one() << 128, causing incorrect limb splitting. Variable names are also updated from limb0/limb1 to high/low to match their semantic meaning, and the output order is corrected to [high, low, U128MulGuarantee]. Test cases are added to verify both the no-overflow case and the case where the product overflows into the high limb.

Type of change

Please check one:

• Bug fix (fixes incorrect behavior)
• New feature
• Performance improvement
• Documentation change with concrete technical impact
• Style, wording, formatting, or typo-only change

Why is this change needed?

BigInt::one().pow(128) evaluates to 1 (one raised to the power of 128), not 2^128. This meant the div_rem was dividing by 1, producing a remainder of 0 and a quotient equal to the full product — making the high/low limb split completely wrong for any non-trivial multiplication.

What was the behavior or documentation before?

The simulation of u128_guarantee_mul produced incorrect high and low limb values due to the division being performed against 1 instead of 2^128.

What is the behavior or documentation after?

The simulation correctly splits the 256-bit product of two u128 values into a high 128-bit limb and a low 128-bit limb using a bitshift (<< 128) for the divisor, and returns them in the order [high, low, U128MulGuarantee].

Related issue or discussion (if any)

N/A

Additional context

U128MulGuarantee is also registered in the bijective type mapping in test_utils.rs to support the new simulation tests.