Throw a KernelError for non-singleton type arguments

[maleadt]

On Julia 1.13 and earlier, Core.Compiler.isconstType(Type{T}) is only true when T has a unique representation. That excludes unions and UnionAlls, so a kernel argument like Type{Union{Missing, Bool}} or Type{Vector} falls through check_invocation to fieldcount, which throws:

ERROR: ArgumentError: type does not have a definite number of fields
  [1] fieldcount(t::Any)
  [2] check_invocation(job::GPUCompiler.CompilerJob)

Such an argument can't work there anyway, because Julia's codegen passes it as a boxed host pointer. This PR reports it as a KernelError:

KernelError: passing a non-singleton type argument

Argument 3 to your kernel function is the type Vector, which
cannot be passed to a GPU kernel on this version of Julia.
Pass `Val(Vector)` instead, or a value of that type.

On Julia with Core.TypeEgal (JuliaLang/julia#62001), methodinstance already turns these into TypeEgal{T} singletons, so the check never fires there. The test covers both cases.

This came up in JuliaGPU/KernelAbstractions.jl#791, where the POCL back-end's zeros(::POCLBackend, Union{Missing, Bool}, ...) passed the element type to a kernel. That is fixed on the KA side in JuliaGPU/KernelAbstractions.jl#795.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.41%. Comparing base (8496909) to head (bf13c3e).
⚠️ Report is 4 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #954   +/-   ##
=======================================
  Coverage   86.41%   86.41%           
=======================================
  Files          29       29           
  Lines        5674     5676    +2     
=======================================
+ Hits         4903     4905    +2     
  Misses        771      771           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.