Drop MATLAB interface

[dweindl]

Following up on #2475, let's get rid of the MATLAB interface which was dragged along, but not really well maintained over the last years.

• port any matlab examples that are still valuable for testing to Python and add them to them CI workflows (Port matlab test models to Python #2898, Start porting MATLAB-based tests to Python #2822, Run full model code generation for test models during CI from Python  #854)
• remove matlab code for model import (Drop MATLAB-interface #2932)
• remove matlab examples (Drop MATLAB-interface #2932)
• remove pre-generated c++ code for matlab test models (Drop MATLAB tests #2931)
• remove matlab interface from the documentation (Drop MATLAB-interface #2932)
• ... as well as any other mentions of the matlab interface in the docs (Remove MATLAB leftovers #2961)
• remove all mtoc / matlab-doxygen related stuff (Drop MATLAB-interface #2932)
• remove matlab GHA workflows (Drop MATLAB tests #2931)
• remove matlab c++ interface code (Drop MATLAB-interface #2932)
• remove all matlab- vs python-code generation switches from the library (Remove MATLAB-specific model code #2957)
• remove Python tests for matlab-pregenerated models (Drop MATLAB tests #2931)
• remove old C++ spline code and obsolete parts from symbolic_functions.{h,cpp} (started in Drop MATLAB-interface #2932, Clean up symbolic_functions.{cpp,h} #2956)
• Update https://raw.githubusercontent.com/AMICI-dev/AMICI/main/doc/gfx/amici_workflow.png (doc: Update workflow figure #2962)

Prerequisites for porting all old tests:

• Requesting sensitivities for algebraic equations breaks simulation #2102
• Adjoint sensitivities for Python-imported models with events #1539
  * [ ] Support for 2nd-order sensitivities #2905 (out of scope)

[dweindl]

• port any matlab examples that are still valuable for testing to Python and add them to them CI workflows

Note that not all of the current Matlab/C++ tests are testing what they are supposed to test.
E.g. TEST(ExampleSteadystate, SensiFwdNewtonPreeq) does not use Newton's method. Probably because the default steady state method was changes to integration-only, but the test relied on the default. And the test code never checked whether Newton's method was used successfully (or at all).

EDIT: At least Newton tests are addressed in #2898.