[NB] support for adjoint derivatives

[dreivmeister]

Purpose

Introduces jacobianSymbolicAdjoint in the NB which generates equations to efficiently evaluate the ODE Jacobian row-wise and not column-wise i.e. compute y = J^T * x and not y = J * x.
This is needed for example in FMI 3.0 (https://fmi-standard.org/docs/3.0/#partial-derivatives).
The current focus is on robust support for scalar models.

Approach

The main changes were done in NBDifferentiate.mo and NBJacobian.mo where NBDifferentiate contains an implementation of reverse-mode AD done symbolically on the Strong Components.