skip IDIC if the generation problem returns infeasible solution

[yuxies]

When dealing with problems featuring lower-level coefficients with a large number of significant digits, the cut generation problem for IDIC might yield an all-zero "optimal" solution that is infeasible due to numerical issues.

• In the future, we may consider checking the greatest common divisor of the coefficients in the lower-level objective function and rescaling them before solving the problem?