Vetted pulse template: timesteps_all_equal=true is uniform-free-duration, not fixed T — switch to pinned Δt_bounds (re-vet flag)

[aarontrowbridge]

Important

Problem — The produced single-qubit template (scores/pulse-designer/templates/solve.jl) constructs its problem with the all-equal option believing it fixes the user-specified gate time T. Measured semantics (Piccolo.jl#317): all-equal enforces a uniform mesh but leaves one free duration-scale degree of freedom — the template's solves are uniform-mesh-free-duration, not fixed-T. The template intends fixed T (the interview asks for T in ns).

Approach — Replace the all-equal option with pinned timestep bounds (Δt_bounds = (T/(N-1), T/(N-1))) so the template means what the interview promises. Behavior change → re-vet flag: the template's vetted fidelities were obtained under the free-scale semantics; re-run the vetted gate set against the pinned landscape before shipping (stalls should improve — the drift-dominated benchmark stalls worse under free-scale).

Scope — in: the one template construction line + its re-vet run. out: other templates, the scores' interview logic, min-time flows.

Acceptance Criteria

• Template construction uses pinned timestep bounds equal to the nominal step
• The vetted gate set re-run and recorded against the changed landscape (fidelities recorded, no regression below the vetted floor)
• Release note per the compatibility policy

Source

• Free-time semantics: safe defaults for unbounded per-knot Δt, three-mode guidance, derived-dt track Piccolo.jl#317 · fix(problems): default Δt bounds window + three-mode free-time guidance DirectTrajOpt.jl#145