nerdamer.divide

This restores the parser-facing compatibility behavior historically exposed as divide(a,b): a successful polynomial division is returned as quotient + remainder/divisor, so the result is a reusable symbolic expression. When more than one variable is present, Nerdamer treats the input as a multivariable polynomial rather than silently choosing a main variable. Inputs outside the polynomial domain fall back to ordinary symbolic division. Once both inputs are polynomial, failures from the division engine are allowed to propagate.