Native/Masm divergence for `u128::saturating_(add|sub)`

### Packages versions

N/A

### Bug description

Adding test coverage for saturating arithmetic fails for `u128` due to a divergence between Rust and Miden VM outputs

### How can this be reproduced?

```
cargo make test -E 'package(midenc-integration-tests)' -- end_to_end::arithmetic::saturating_add_u128
```

### Relevant log output

```shell
thread 'end_to_end::arithmetic::saturating_sub_u128' (215234908) panicked at tests/integration/src/end_to_end/arithmetic.rs:649:13:
Found minimal(shrinked) failing case: (340282366920938463463374607431768211455, 340282366920938463463374607431768211455)
Failure: Reason("assertion failed: `(left == right)` \n left: `340282366920938463444927863358058659840`, \n right: `0`: the Miden VM and Rust disagree on the value produced by this operation at tests/integration/src/end_to_end/arithmetic.rs:638")
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
```

The failing test is in tests/integration/src/end_to_end/arithmetic.rs, line 638. Run the specific test with the given cargo command to see the divergence between Rust and Miden VM for u128 saturating operations. Examine the compiler's code generation for saturating arithmetic on u128, likely in the miden backend or intrinsic handling. The goal is to make the Miden VM output match Rust's saturating_add and saturating_sub for the max value inputs shown.

由索引模型根据 Issue 内容生成。