Nested der-expressions generates invalid simulation code

[Per Östlund]

Nested der-expressions seems to cause Exp.solve to fail, which causes generation of simulation code that doesn't compile. Example model:

class MyRealEquation
  Real x(start=0);
equation
  100 - 20*der(x) + der(x)*der(x) = 40*der(der(x));
end MyRealEquation;

[Per Östlund]

Still an issue, although the compiler now gives an internal error saying that derivates of order > 1 is unsupported. It still continues to generate simulation code though, which of course does not compile. Dymola doesn't support derivates of order > 1 either, so it's probably fine to not support it. Maybe it's not even allowed in Modelica, I don't know. But the compiler should show a proper error in that case, and stop the compilation.

[Martin Sjölund]

It should be possible to handle this (compile it), but not without getting a warning that some introduced dummy variable der_x not having specified initial conditions.

[Henning Kiel]

Just another note: if you change der(der(x)) to something like der(x*der(x)) the nested der is not detected, i.e. the internal error is not raised. Compilation, of course, still fails.

[Henning Kiel]

Changing BackendDAEOptimize.expandDerExp to use match instead of matchcontinue makes omc stop compilation at this error, but I cannot estimate the impact on other working models. Shall I create a PR?