Last attempt at using hand-built Newton-Raphson solver

About to mothball the simple_classes branch. It uses the NR solver in solver.cpp which works OK when the projection is well behaved, but had problems when imposing bounds on effort.

PAW 2025 version

Used to generate figures and tests for PAW 2025

Mix catch and effort targets

Works with test script yft_projection_solve_test.R

Catch target just about working

Just about works with a catch target. Tagging it so I know which script works with this version before I make a hash of everything.