Use QDigit in square algorithm if it is available

[OTheDev]

Currently, the multiple-precision squaring algorithm used needs space for 3 times the width of a Digit and this is achieved by using two double digits DDigit to act as one QDigit.

When Digit::BITS is 32 bits or less, an unsigned integer type with width equal to 4 times the width of a Digit exists: QDigit.

This PR updates the algorithm to use QDigit explicitly if it is available, rather than implement QDigit on the fly using two DDigits.

This improves maintainability and potentially, efficiency.