Function lexical_util::div128::slow_u128_divrem

pub fn slow_u128_divrem(n: u128, d: u64, d_ctlz: u32) -> (u128, u64)

Optimized fallback division/remainder algorithm for u128.

This is because the code generation for u128 divrem is very inefficient in Rust, calling both __udivmodti4 twice internally, rather than a single time.

This is still a fair bit slower than the optimized algorithms described in the above paper, but this is a suitable fallback when we cannot use the faster algorithm.