Updates the requirements on [CUDA](https://github.com/JuliaGPU/CUDA.jl) to permit the latest version.

Updates `CUDA` to 6.0.0
- [Release notes](https://github.com/JuliaGPU/CUDA.jl/releases)
- [Commits](JuliaGPU/CUDA.jl@v5.0.0...v6.0.0)

---
updated-dependencies:
- dependency-name: CUDA
  dependency-version: 6.0.0
  dependency-type: direct:production
  dependency-group: all-julia-packages
...

Signed-off-by: dependabot[bot] <support@github.com>

…s-629bbb4d39

Update CUDA requirement from 5 to 5, 6.0 in the all-julia-packages group across 1 directory

`[zero(u) for u in VA.u]` always produces a plain `Vector`, even when
`VA.u` is a `StructVector` (from StructArrays.jl). The constructor
then fails with `MethodError: Cannot convert Vector{SVector{1,Float64}}
to StructVector{...}` because the type parameter `A` is locked to the
original container type.

Use the existing `rewrap(parent, u)` mechanism (already used by
broadcast) to convert the comprehension result back to the original
container type. The `StructArrays` extension defines
`rewrap(::StructArray, u) = StructArray(u)`, and the fallback is
`convert(typeof(parent), u)`.

Fixes OrdinaryDiffEq.jl v7 CI failures in OrdinaryDiffEqLowStorageRK
and OrdinaryDiffEqSSPRK "VectorOfArray/StructArray compatibility" tests.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Verifies that the rewrap-based fix correctly returns a VectorOfArray
whose .u field is still a StructArray (not a plain Vector) and that
all values are zeroed.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Preserve container type in zero(VectorOfArray) via rewrap

Update Project.toml