Replace cumbersome A matrix stacking logic for linear constraints with a single ObjectiveFunction for all linear objectives

[dpanici]

Action items @ddudt

• create ObjectiveFunction inside of optimizer.optimize that takes in the linear constriants
• call combine_args on it and objective before we do the linear constraint projection
• now factorize_linear_constraints argument self._constraints in (

  DESC/desc/optimize/_constraint_wrappers.py

  Line 94 in 3707e08

  ) = factorize_linear_constraints(

  ) is now just an objective function, use its jac_scaled argument to get A_full in (

  DESC/desc/objectives/utils.py

  Line 102 in 3707e08

  A_full = jnp.vstack([jnp.hstack(Ai) for Ai in A])

  ) and call obj.compute(zeros) to get b_full
• keep the indexing stuff after that line above, and everything should hopefully work fine for optimizable optimization (not yet for OptimizableCollection)