Use OCP functions true sparsity patterns in manual mode

Will be easier after https://github.com/control-toolbox/CTDirect.jl/issues/417

This should reduce the number of nonzero elements significantly, without adding much to the cost of computing the manual sparsity patterns (hopefully)

- [ ] Utils: discretization dependent getters that return the indices range for specific variables (x_i, u_i, v) ? Would be useful to set the nonzero blocks for the sparsity patterns of each OCP function. These can also be reused by the existing getters for the variables. Start with Euler since the patterns are the simplest ?

- [ ] Hessian (likely the most impactful)
- [ ] mock-up calls to the OCP functions sparsity patterns (use full nonzero for now)
- [ ] dynamics
- [ ] lagrange cost
- [ ] path constraints
- [ ] mayer cost
- [ ] boundary conditions
- [ ] retrieve and use actual sparsity patterns of order 2

- [ ] Jacobian