Fix BigInt#% for unsigned integers

[straight-shoota]

Fixes #12770

[jzakiya]

Thank you!

[asterite]

Thank you! There's another such occurrence in the // method.

[straight-shoota]

That's uncritical because unsigned types won't reach this def (there's a //(other : Int::Unsigned) overload).

[straight-shoota]

But there is indeed a common pattern of this form:

if other < 0
  (-self).unsafe_op(-other)
else
  unsafe_op(other)
end

It would be more concise to express this as something like (other.sign * self).unsafe_op(other.abs).
Performance should probably be similar in release mode.

[HertzDevil]

This still breaks if other is e.g. Int32::MIN, but we can worry about that later because it's not just this method