Feature: Better tools for solving ODEs

[flo-schu]

It is inconvenient that model parameters of ODE-systems have to be re-specified under the current implementation (i.e. when calling the solver). :warning: Note that the tools should be at most tools that pymob offers. They should not be interconnected with the core-pymob functionality

Desired features:

• [x] inspect model function call to get all arguments that are not t and X. The remaining parameters should be model parameters. These can be then be matched to order the values of an incoming parameter dictionary to provide *args for the solve_ivp call 🙂
• [x] ⚠️ Fix seeding problem. pathos uses pickling to serialize objects --> this results in a copied seed queue in parallel model evaluations, meaning that seeds are reused for stochastic simulations, ideally, seeds are passed when simulations are launched.
  • [x] This could be attacked by using an Evaluator class, which is instantiated each time the simulation is evaluated
  • With this also the deprecation warning for using the results property (which is convenient) inisde the simulation class can be avoided.

[flo-schu]

• 7 has been in the center of this pull request and is completed for the reversible_damage case-study.

When testing of the numpyro backend is completed. The branch can be merged