Event support in integrator class

[jaeandersson]

With the improving support for Modelica (#3670, #3671) and FMI (e.g. #3160), it becomes natural to model hybrid systems, i.e. systems with events dynamics. In order to simulate such systems in CasADi, the integrator class needs to be generalized.

Implementation steps:

• Add a zero-crossing field to the DAE oracle function
• Add an option in the integrator class corresponding to the maximum number of events allowed (e.g. 20). Memory will need to be allocated for each event. If the number of events is exceeded during integration, an error will be thrown
• Add an option to the integrator class corresponding to a transition function for the state and guess for algebraic variables. When a zero crossing (from negative to positive) happens, the transition function will be called, updating the state and algebraic variables accordingly. If no function, assume continuous state trajectory. Event handling can still be useful, e.g. if there is a discontinuity in the ODE right-hand-side.
• During integration, have the integrator class project at what time the next event will happen by linear extrapolation. That is, before the integration starts for a particular control/output interval, the value for each zero crossing function is calculated along with its time derivative. If a particular zero crossing function is negative and its time derivative is positive, we can get an estimate of then the event will occur. If any event is expected to happen before the end of the interval, adjust the end time to only integrate to that time.
• If an "event step", iteratively find the event time using linear extrapolation (should converge much faster than bisection). Make sure that the no other zero-crossing happened before it (in which case the procedure needs to be repeated for that other zero-crossing) and that the output time wasn't reached before the event happened.
• Even if no event, we should still check if an event occurred as the linear extrapolation might have missed it
• Numeric forward sensitivity analysis support
• Numeric adjoint sensitivity analysis support
• Support for forward sparsity pattern propagation
• Support for reverse sparsity pattern propagation
• Symbolic forward sensitivity analysis support
• Symbolic adjoint sensitivity analysis support