#244 Polynomial operators do not respect accuracy for backwards-propagation

[Ulf Lorenz]

The cut-off for the Chebychev or Faber solvers is always based on the coefficients for forward-propagation, but applies for both coefficients (forward and backward propagation). The correct solution would be to cut-off the forward and backward coefficients separately.

Low priority. Does anybody use propagation backwards in time for time-independent systems?