Returns the number of significant digits that suffice to losslessly represent any Float of
precision prec in base base: printing such a Float to this many digits with rounding to
nearest (for example with get_str), then reading the digits back at precision prec, again
rounding to nearest, recovers the original value exactly.