A way to combine variable-step integration with a discrete-time controller

[sawyerbfuller]

Perhaps using a different integrator than the default (e.g. LSODA?) in scipy.integrate.solve_ivp?

The example in https://github.com/python-control/python-control/blob/main/examples/simulating_discrete_nonlinear.ipynb is one option, but it uses a fixed-step integrator (edit: which is not as accurate as a variable step integrator and potentially slower). Constructing discrete-time systems that repeatedly call ``scipy.integrate.solve_ivp``` inside the dynamics function is another option, but it is very slow.

[murrayrm]

I've thought about this a bit in the past (though done nothing about it -:). One possibility might be to use the events parameter in solve_ivp, which could be set up to create events whenever a discrete time "tick" occurs. Another would be to call continuous time integration between discrete clock ticks (similar to the discussion in #933). I'm not sure whether either of those are faster than using a fixed-step integrator.