fix write_solution for vector fields

[marinlauber]

This PR fixes the write_solution crashes for vector-valued fields (e.g. 3D displacement Lagrange{RefHexahedron,1}()^3) with the error: number of base functions (8) does not match length of vector (24)

The issue is that _evaluate_at_grid_nodes! contains a type check to decide whether to reinterpret the local dof vector as a Vec:

function Ferrite._evaluate_at_grid_nodes!(data::Union{Vector,Matrix}, dh::NODDofHandler, sdh::SubDofHandler,
        u::Vector{T}, cv::CellValues, drange::UnitRange, ::Type{RT}) where {T, RT}
        ...
        if RT <: Vec && cv isa CellValues{<:ScalarInterpolation}
            uer = reinterpret(RT, ue)
        else
        ...
end

However, CellValues is parameterised as CellValues{FV<:FunctionValues, GM, QR, detT} without the interpolation is no longer in the type parameter. The isa check therefore always returns false for vector fields, so uer = ue (length 24) is passed to function_value(cv, qp, uer) which expects length 8.

The fix is very simple and only requires changing the check to RT <: Vec && Ferrite.function_interpolation(cv) isa Ferrite.ScalarInterpolation.

This PR was co-authored by Claude Code v2.1.146 (Sonnet 4.6).

PS. I'm happy to add my example with vector field to the docs if you want? (Neohookean cube under prescribed extension rotation with NR solver.)

[termi-official]

Thanks for tackling this Marin!

Yes, that would be indeed an interesting addition to the docs here to show the Newton-Raphson in distributed.