Implement FBA constraint iteration until new pathway is found

[AlecGower]

Currently when using pathways to constrain FBA, if the FBA simulation returns infeasible or a growth below the threshold, we mark that the pathway in question was infeasible.

Really, that pathway provided is just one possible pathway of potentially very many for the given genotype and conditions.

What this enhancement suggests is that in the case of an infeasible result from FBA constraint, we should instead:

1. Systematically remove reactions from the list of constrained ones until growth is restored (in the FBA simulation).
2. Remove reaction from the constraint set.
3. Conduct FBA simulation
4. Restore the last removed reaction if FBA is now feasible.
5. Proceed down the list of reactions
   This will result in "minimal restrictive pathway", in a number of FBA simulations equivalent to the number of reactions found in the original pathway (i.e. a subset of reactions that should be avoided by LGEM+ when deducing pathways).
6. Add this set as a disjunction of negations to the original theory files.
7. Identify if there is a pathway now after the deletion.
8. Repeat!

Some additional notes in my notebook and in Konstantin's comments on the manuscript from 22 March.