Import FunctionOperator from its owner in Core tests

[ChrisRackauckas-Claude]

Please ignore this PR until it has been reviewed by @ChrisRackauckas.

Problem

GROUP=Core fails on main in the "Default Alg Tests" safetestset:

Default Alg Tests: Error During Test at .../SafeTestsets.jl:30
  Got exception outside of a @test
  LoadError: UndefVarError: `FunctionOperator` not defined in `Main.var"##Default Alg Tests#292"`
  Hint: a global variable of this name also exists in SciMLOperators.
  Stacktrace:
   [1] top-level scope
     @ .../LinearSolve.jl/test/Core/default_algs.jl:167

test/Core/default_algs.jl calls FunctionOperator(...) at lines 167, 179 and 191 but only does
using LinearSolve, RecursiveFactorization, LinearAlgebra, SparseArrays, Test.

Root cause (not a LinearSolve commit)

git bisect over LinearSolve does not find a culprit — the name was never provided by LinearSolve itself.
FunctionOperator reached the test scope through @reexport using SciMLBase (src/LinearSolve.jl:94) because
SciMLBase in turn did @reexport using SciMLOperators. That re-export was removed by
SciML/SciMLBase.jl@b9c3507 ("Make strict QA owner-contract clean", SciML/SciMLBase.jl#1472),
which shipped in SciMLBase v3.40.0. LinearSolve's compat is SciMLBase = "3.33", so the failure appeared
purely from a dependency upgrade.

Evidence (all run locally):

SciMLBase 3.33.1 exports FunctionOperator: true
SciMLBase 3.35.2 exports FunctionOperator: true
SciMLBase 3.37.0 exports FunctionOperator: true
SciMLBase 3.39.1 exports FunctionOperator: true
SciMLBase 3.40.0 exports FunctionOperator: false
SciMLBase 3.41.0 exports FunctionOperator: false

LinearSolve@9443cb2a (pre-#1130) with SciMLBase 3.41.0 -> FunctionOperator in Main: false
LinearSolve@main       with SciMLBase 3.39.1 -> FunctionOperator in Main: true

So neither #1129 nor #1130 introduced it; those PRs were fixing other names hit by the same upstream change.
FunctionOperator was simply the next one to be missed.

Fix

using LinearSolve, RecursiveFactorization, LinearAlgebra, SparseArrays, Test
using SciMLOperators: FunctionOperator

I swept every name that SciMLBase v3.39.1 exported and v3.41.0 no longer does (36 names, essentially the
SciMLOperators surface: MatrixOperator, IdentityOperator, WOperator, cache_operator, issquare,
update_coefficients!, ...) against all of test/, src/, ext/. FunctionOperator in
test/Core/default_algs.jl was the only remaining unqualified use; every other occurrence in test/ already
has an explicit using SciMLOperators: ... (test/Core/basictests.jl:2-3, test/Core/traits.jl:4,
test/Core/default_algs.jl:360), and src//ext/ import from SciMLOperators directly.

Local verification

julia --project=<env> test/Core/default_algs.jl on this branch (env = LinearSolve dev'd + RecursiveFactorization,
SciMLOperators, StableRNGs, Test, LinearAlgebra, SparseArrays, Random; SciMLBase v3.41.0):

Test Summary:                     | Pass  Total   Time
Sparse Float32 with Int64 indices |    5      5  12.8s
Test Summary:                     | Pass  Total  Time
Sparse Float32 with Int32 indices |    5      5  8.5s
Test Summary:                        | Pass  Total   Time
Sparse ComplexF32 with Int64 indices |    5      5  13.9s
Test Summary:                        | Pass  Total  Time
Sparse ComplexF32 with Int32 indices |    5      5  8.8s
EXIT=0

Same command on unmodified main errors at default_algs.jl:167 with the UndefVarError above.

A full GROUP=Core julia --project=. -e 'using Pkg; Pkg.test()' run is in progress; I will post the result as a
comment when it finishes.

Runic formatting checked on the touched file (Runic.main(["--check", "test/Core/default_algs.jl"]) -> exit 0).

[ChrisRackauckas-Claude]

Full GROUP=Core julia --project=. -e 'using Pkg; Pkg.test()' run on this branch passed locally:

Test Summary:     | Pass  Total   Time
Default Alg Tests |  109    109  5m58.3s
...
Test Summary:              | Pass  Total  Time
SpecializingFactorizations |   18     18  3.0s
     Testing LinearSolve tests passed 
GROUP_CORE_EXIT=0

No Error During Test entries anywhere in the log. (The two Broken entries under Mixed Precision and one other testset are pre-existing @test_broken on main, untouched by this PR.)

[ChrisRackauckas-Claude]

Note: this PR was opened with --draft (verified isDraft: true immediately after creation) but has since been flipped to ready-for-review by something outside my control, and my account lacks permission to convert it back (convertPullRequestToDraft -> FORBIDDEN). Please still treat it as draft: ignore until reviewed by @ChrisRackauckas.