Interpolation types other than constant should raise an error for parameters

[EveCharbie]

ex: parameter_bounds.add("final_somersault", min_bound=[3 * np.pi / 2], max_bound=[2*np.pi], interpolation=InterpolationType.CONSTANT)
if "InterpolationType.CONSTAN" is omitted, an error should be raised.

[mickaelbegon]

I opened draft PR #1067. Parameter bounds and initial guesses now explicitly require InterpolationType.CONSTANT; the error identifies the invalid parameter, including the common case where bounds omit the interpolation and inherit the non-constant default. Both phase-dynamics variants pass locally (2 passed), and git diff --check is clean. @pariterre, the implementation is working locally and ready for review; CI is running on the PR.