Fix possible error in jacobian tests

[ronald-jaepel]

During the jacobian tests we compare the residual calculated with finite differences in both directions (stored in colA and colB):

... if (std::isnan(colA[row]))
  CHECK(std::isnan(colB[row]));
else
  CHECK(std::signbit(colA[row]) == std::signbit(colB[row]));

However, the first test is:

if (std::abs(colA[row]) <= absTol)
  CHECK(std::abs(colA[row]) <= absTol);

where we compare colA to colA. To me this seems like a typo, because very other check compares the result in colA with colB.

With this change some test fail now.

ToDos:

• someone else confirm that this change is reasonable
• fix tests again
  • GRM_DG multiple particle types Jacobian analytic vs AD
  • GRM_DG dynamic reactions Jacobian vs AD bulk
  • GRM_DG dynamic reactions Jacobian vs AD particle
  • GRM_DG dynamic reactions Jacobian vs AD modified particle
  • GRM_DG dynamic reactions Jacobian vs AD bulk and particle
  • GRM_DG dynamic reactions Jacobian vs AD bulk and modified particle
  • Crystallization Jacobian verification for a DPFR/LRM with primary and secondary nucleation and growth

[jbreue16]

Thanks for catching this !
Turns out we just need to adapt tolerances within a reasonable range :)