Clean-up steady-state solver code

[paulstapor]

Currently, the workSteadyStateProblem routine isn't exactly the nicest part of Amici... Things to be changed here in some close future:

• [x] Remove the triple try-catch nesting, when determining whether or not to use simulation
• [x] Allow steady-state sensi computation via solving the linear system, if steady-state was computed via simulation: Currently, that's not possible and full FSA computation is carried out, which is fatal for large models.
• [x] Add flag "simulationOnly" to steadyStateSensitivityMethod, which does not even try to launch the Newton solver.
• [x] ~#1675~
• [x] Merge #917
• [ ] ~Replace custom implementation of iterative linear solver in Newton Solver by~ Interface GMRES, BICGSTAB, and TFQMR (do it analogously to solver.cpp) (#1164 )
• [ ] correct the (partly) erroneous or misleading output into rdata of the steady state solver
• [ ] create figure for documenting the steady state solver code via sphinx, rather than having something prerendered
• [x] #1737

[dweindl]

* [ ]  Remove the triple try-catch nesting, when determining whether or not to use simulation

Very much in favor. This is a huge pain to debug.

[FFroehlich]

• [ ] correct the (partly) erroneous or misleading output into rdata of the steady state solver

I have no idea what this refers to, anyone?

[dweindl]

No idea.

[dweindl]

* [ ]  create figure for documenting the steady state solver code via sphinx, rather than having something prerendered

Started here. Might need some further updates already. These diagrams can be included in sphinx via https://github.com/mgaitan/sphinxcontrib-mermaid.