96.2.2 Rat: replaced <x> by <y> = <z>

Floating-point numbers provide a maximum number of digits that is typically high, but still limited. Good examples that this limitation might be too low even for harmless-looking examples include Wilkinson’s Polynomial, The Rump polynomial and the fact that an exact 1/10 cannot be expressed as a binary floating-point number. In places where the floating-point error might add up or hinder terms from cancelling each other out maxima therefore by default replaces them with exact fractions. See also ratprint, ratepsilon, bftorat, fpprintprec and rationalize.