Change reaction constraints using un-mangled reaction id

[stelmo]

right now the change constraint modification uses reactions(model) to set bounds on variables. For more complicated models like GeckoModels you can't use the natural (un-mangled) reaction ids to set the bounds on variables. We should fix it so that the users can do this without having to look-up the mangled ones (or changing the underlying model).

[exaexa]

this could basically multiply the constraint through the reaction_fluxes matrix and add it as is. Certainly doable, I'll have a look at it.

OTOH if possible one should preferably set the "base" reaction bounds on the original model before it gets through gecko construction.

[exaexa]

naming: change_constraint vs change_flux_bound ?

(future after rename: change_variable_constraint vs change_reaction_constraint ?)

[exaexa]

solved differently (no mangling in C2)