Dof ordering for vector/tensor-valued fields about gridap.jl

[santiagobadia]

The node major is more similar to what we want (I guess) for multi-field, i.e., a block structure. For instance, the multi-valued Laplacian is a good approximation of the elasticity operator, and using node major, you have a nice block structure.

The component major is needed for AMG methods that in practice do not perform well for multi-component problems, e.g., AMG solvers for compressible elasticity.

If you ask me, I would say node major. In any case, we should probably have the possibility to choose both.

[fverdugo]

Provide the option to chose one or another for the global dofs.

For the local dofs "node major" is already enough