Find the best way to run a flowsheet the user has already chosen.

The topology is fixed: no step is added, removed or reordered. Only parameters that design_space.py classifies as tunable are moved, so the search cannot make the process look cheap by assuming a better resin or membrane than the one being costed.

Returns the optimised parameters per step, ready to write straight back onto the canvas, together with the baseline they beat and a note on any value that landed on a bound.