If I want to convert a Vec<u64> to BigUint, I need to first convert it to Vec<u32> manually and then BigUint::new converts it back to Vec<BigDigit = u64> internally.
From other issues I gather that you don't want to expose BigDigit arguing cross platform compatibility. Can we then add new methods i.e. BigUint::new_u64(digits: Vec<u64>) that would be a zero-copy constructor on 64 bit platforms?