Pathsim

- reworked error controllers for the adaptive timestep integrators
- added GEAR type integrators to the solver library (multistep methods with adaptive timestep and adaptive order selection)

- fixed path length estimation for nested subsystems and added tests and an example
- renamed utils.funcs to utils.utils, which makes more sense

- minor fixes
- fixed RKDP54 butcher table
- test coverage for Value and Parameter classes for AD framework

- refactoring of numerical integrators (solvers), separate parent classes for diagonally implicit and explicit runge kutta methods simplifies the specific methods significantly
- performance of runge kutta methods improved >10% by using explicit loops with indexing to construct slope instead of sum comprehension

- created interface for automatic differentiation framework to numpy ufuncs
- updated testsuite
- general bugfixing regarding implicit solvers (jacobian computation)

Lots of refactors, improvements and new features since the last release including:

- Added rudimentary automatic differentiation (AD) framework in `pathsim.diff` to make the simulations differentiable w.r.t. predefined parameters
- Refactoring of numerical integrators to be compatible with AD framework for transient simulation
- Added lots of test, initial coverage was above 85%, more coming in the future
- Overall cleanup