Refactor Simulation.solve

[brosaplanella]

As part of #3101 I was looking at the Simulation class and noticed that the messy part is basically the solve method (and in particular, the experiment part). In my mind, the bigger goal here is to keep Simulation class (without experiments) to be compatible with any model (even models beyond battery).

I think that just simply refactoring the whole experiment part as a separate function would help a lot (maybe with an extra check of whether a given model supports experiments or not). The other main point I would like to discuss here is whether we deprecate the "drive cycle" mode, given that now this can easily be done with the experiment steps.

TLDR What I propose here is:

• Get rid of the "drive cycle" mode in the Simulation (already supported via experiment).
• Refactor the code that solves the simulation for experiment.
• Ensure the "without experiment" mode works with any model (even those custom defined).
• As part of this, refactor experiment to better handle different start/end times for identical steps.

[brosaplanella]

I have noticed that this would probably require defining self.callbacks and self.logs. I am not familiar on how they work at all, so would there be any issue with this?

[brosaplanella]

I have noticed that this would probably require defining self.callbacks and self.logs. I am not familiar on how they work at all, so would there be any issue with this?

After discussion in the developers meeting that should be fine.