[Particles] Introduce constraints feature to the particles framework

[vovannikov]

To proceed with the 2D simulations in the SPH setting, we need to ensure that no action is produced in the z-direction, that may appear when the particles are coupled to 3D structures, e.g. particle walls. It is then required to explicitly nullify the corresponding velocity and acceleration components, since the 2D SPH kernel is not capable to handle these contributions properly.

As originally proposed by @slfuchs, such sanitation is easy to perform in the pre_interaction_routine() step of the time integrator. I just decided to introduce the concept of constraints, that would encapsulate this functionality and that could be extended to other tasks if needed. One could use the Dirichlet boundary conditions handler, but that handler has a bit more functionality than needed, making it a bit too expensive to apply at each timestep to each particle of the simulation domain.

Together with #1728, the 2D SPH cases requiring coupling to structural elements can not be simulated (including those from the paper of @slfuchs).

[ppraegla]

Thanks for the implementation. Could you add an input file test that uses the constraints? Right now, create_constraints is unused (unless I missed something).

[slfuchs]

I like the idea of introducing a constraints handler

[vovannikov]

After thinking further and also looking at #1728, I could extend the flexibility of this feature and detect to which plane (2D) or axis (1D) projection should take place. This will be also merged with the coordinates check @slfuchs mentioned.

[vovannikov]

Thanks for the implementation. Could you add an input file test that uses the constraints? Right now, create_constraints is unused (unless I missed something).

Done! Now there is a test that crashes without this feature.

[vovannikov]

Here, for the test to pass, #1728 needs to be merged first. Once #1728 is merged, I will rebase this one and proceed.

Ready for review.

As I have commented above, I would like to remove the requirement for the particles to be located at the plane z=0. I have already programmed an algo that detects the plane automatically and checks whether all the particles are coplanar, but I would prefer to add it in a follow up PR due to its relative complexity. Again to keep things for review simple.