`cons_jac_prototype` sometimes ignored for sparse backend

[baggepinnen]

When providing a cons_jac_prototype but not a Jacobian function, the provided prototype has no effect and is ignored.

I have a function for which SCT fails but it is easy to provide the prototype, so it would be nice if this worked

[baggepinnen]

It seems the way to make use of this is to load ADTypes and provide

backend = AutoSparse(DenseBackend(); sparsity_detector=ADTypes.KnownJacobianSparsityDetector(cons_jac_prototype))