fix: correctly parse small floats, use correct comparison in add

[zbaylin]

This fixes two bugs:

• When serializing small floats, string_of_float uses scientific notation which the current parser doesn't account for. Furthermore, there is no way to specify the amount of precision in the resultant string.
  • Solution: use Printf.sprintf to serialize the string with the precision being context.prec
• When adding numbers, we check to see if they are equal but of opposite sign. This is done using string comparisons which do not take into account leading 0s that can be added by normalize
  • Solution: parse the coefficients as Z.ts earlier and use those for the comparison