Fix DefaultsLoading test: sparse BigFloat defaults to PureKLU (not Sparspak) since #1037

[ChrisRackauckas-Claude]

Fixes #1042 — a stale test failing on main/CI in the DefaultsLoading group.

Problem

test/defaultsloading/defaults_loading.jl:30 asserted, via @test_throws, that solving a sparse BigFloat LinearProblem before using Sparspak raises "SparspakFactorization required ... using Sparspak". That fails on main with "No exception thrown":

Defaults Loading Tests: Test Failed at .../defaults_loading.jl:30
  Expression: sol = (solve(prob)).u
    Expected: ["SparspakFactorization required", "using Sparspak"]
  No exception thrown

Cause

#1037 (52ebf258, "Default to PureKLU for generic-eltype sparse LU") intentionally changed the default sparse LU for generic (non-BLAS) element types from SparspakFactorization to PureKLU — a pure-Julia hard dependency — and removed the "SparspakFactorization required" error branch from defaultalg. So that error path is unreachable now, and the @test_throws never fires. The test simply wasn't updated alongside the routing change.

Fix

Update the test to the new intended behavior: a generic-eltype (BigFloat) sparse problem routes to KLUFactorization and solves without using Sparspak.

@test Base.get_extension(LinearSolve, :LinearSolveSparspakExt) === nothing
@test LinearSolve.defaultalg(mat, rhs).alg ===
    LinearSolve.DefaultAlgorithmChoice.KLUFactorization
sol = solve(prob).u
@test sol isa Vector{BigFloat}

Verification

Ran the full test/defaultsloading/defaults_loading.jl locally (Julia 1.12 / Linux, with Sparspak + STRUMPACK_jll available) — passes, no failures. Runic --check clean.

---

Please ignore until reviewed by @ChrisRackauckas.

🤖 Generated with Claude Code