Hi, @santiagobadia,

This PR adds (generalised) Modal C0 bases and reffes into Gridap.

I have refactored the code such that it is minimally invasive.

There are two things however to improve:

1. Modal C0 reffes are generally different at each cell, because they express different shape functions according to the cell bounding box. In the implementation, I am assigning a different Modal C0 cell type to each cell. On the other hand, the cell-wise ownership given by CellConformity is compressed on cell types. At the moment, I compute the CellConformity for each Modal C0 cell type. This is redundant. We have different shape functions, but the CellConformity is usually the same everywhere. I have found out that this approach is quite inefficient in

   Gridap.jl/src/FESpaces/ConformingFESpaces.jl

   Line 78 in 1fb3dc9

   function CellConformity(cell_reffe::AbstractArray{<:ReferenceFE},conf::Conformity)

   and

   Gridap.jl/src/FESpaces/ConformingFESpaces.jl

   Line 40 in 1fb3dc9

   function _d_ctype_ldface_own_ldofs(a::CellConformity)

   We should definitely compute once (the shared) CellConformity and assign it to all Modal C0 cell types.
2. In order to reuse the procedure that creates the AgFE space on top of the standard one, I am expressing the Modal C0 shape functions as a trivial linear combination in

   Gridap.jl/src/ReferenceFEs/ReferenceFEInterfaces.jl

   Line 491 in eadad31

   linear_combination(Eye{Int}(ndofs),shapefuns)) # Trick to be able to eval dofs af shapefuns

   If I don't do this I cannot evaluate dofs at root shape functions in the physical domain, as Gridap errors at attempting to evaluate a single Modal C0 shapefun at a single point. I think Gridap's design is planned to evaluate the whole Modal C0 basis at a single point, not a single shapefun. Maybe @fverdugo knows what is missing.

We can discuss about those two points whenever you want.

In the meantime, the new functionality (up to the points above) can be reviewed and merged into the code.

Cheers!