[HOLD] Reject all-points control initial guesses

[mickaelbegon]

Summary

• reject InterpolationType.ALL_POINTS for control initial guesses
• report the unsupported interpolation before optimization-vector construction
• update the initial-guess regression test for all randomization and phase-dynamics variants

Root cause

NonLinearProgram.update_init validated ALL_POINTS for states and algebraic states but omitted controls. Collocation sub-points exist for state decision variables, not control decision variables, so accepting this interpolation for u_init is ambiguous and can lead to inconsistent dimensions.

Validation

• MPLCONFIGDIR=/private/tmp/bioptim-mpl-970 python -m pytest tests/shard2/test_global_getting_started.py -k 'test_initial_guesses and ALL_POINTS' -q (4 passed)
• git diff --check

Closes #970

---

This change is