Field dim is not necessarily spatial dim

[koehlerson]

When creating *VectorValues it is assumed that the field has the same dimensions as the spatial ones, which is not always the case, e.g. when evaluating gradient fields

[termi-official]

Similar for the other dimensions. Reference element dimension is not necessarily spatial dimension and field dimension is also decoupled from vectorization dimension of scalar valued basis functions.

E.g. Nedelec has field dimension D, but only one basis function, wheras vectorized Lagrange has field dimension D with D basis functions per entity.

[termi-official]

Relates to #569 and #398

[fredrikekre]

Fixed by #694 + #651 I think? Reopen/comment if not.

[termi-official]

Not fully fixed yet. This code still fails

using Ferrite
grid = generate_grid(Quadrilateral, (2, 2))
dim = 2
ip = Lagrange{RefQuadrilateral, 1}()
qr_face = FaceQuadratureRule{RefQuadrilateral}(2)
facevalues = FaceValues(qr_face, ip^3);

because the embedded values was mostly about getting CellValues in place. But should be easy to fix.

Edit 1: Marked the remaining fix as a good first issue. Basically we have to mirror the functionality of the new CellValues constructors into FaceValues.

Edit2: New interface.