Port MTK test constructions from deprecated ODESystem to System

[ChrisRackauckas-Claude]

Summary

ModelingToolkit deprecates ODESystem(args...; kwargs...) in favor of System(...), so GROUP=CPU tests abort under --depwarn=error on the two remaining ODESystem calls in test/gpu_kernel_de/stiff_ode/gpu_ode_modelingtoolkit_dae.jl (the "Host symbolic setter" and "MTK Pendulum DAE" testsets). This PR swaps both to System with identical equations, parameters, and defaults — the rest of the file already uses System.

Verification

Machine has no NVIDIA GPU; verified on the CPU backend only.

Red before (on master, 4f726d5+b6ae953):

GROUP=CPU julia --depwarn=error --project=test -e 'include("test/gpu_kernel_de/stiff_ode/gpu_ode_modelingtoolkit_dae.jl")'
...
Host symbolic setter with trivial initialization: Error During Test
  `ODESystem(args...; kwargs...)` is deprecated. Use `System(args...; kwargs...) instead`.
ERROR: LoadError: Some tests did not pass: 0 passed, 0 failed, 1 errored, 0 broken.

Green after (this branch, same command):

Test Summary:                                    | Pass  Total   Time
Host symbolic setter with trivial initialization |    5      5  26.8s
Test Summary:                        | Pass  Total   Time
MTK Pendulum DAE with initialization |   21     21  26.9s
...
Test Summary:                              | Pass  Total  Time
make_static_storage is the conversion hook |    7      7  0.2s

Every testset in the file (84 tests total; CUDA CI on this PR reports the same 84) passes; exit code 0. runic --inplace made no further changes and typos is clean on the file.

Not verified

• CUDA/AMDGPU/JLArrays/OpenCL runs of this file (no GPU on this machine); the change is inside @testset bodies that run identically on every backend, and the file already uses System in four other testsets.
• docs/src/tutorials/modelingtoolkit.md still spells the same deprecated ODESystem in an @example (only exercised by the docs build on a GPU runner) — left alone to keep this PR test-only.
• ext/CUDAExt.jl still calls deprecated CUDA.CUBLAS.getrf_strided_batched! (warning seen during QA precompile; CUDA-only path, not hit on CPU). Reported, deliberately not changed here.

Please ignore this draft until reviewed by @ChrisRackauckas.

Risk assessment

• Risk: Low — two-token rename inside test-only code; System is a strict rename of ODESystem in current ModelingToolkit and is already used elsewhere in the same file.
• Blast radius: One test file; no src/ext changes.
• Evidence: --depwarn=error failure reproduced on master and shown passing on this branch (CPU backend).
• Independent review: Claude Code (claude-opus-5-5) rated it low risk, high confidence, verdict MERGE. It reran the file (84/84) and checked the CUDA CI on this PR. Note: master CI runs with --depwarn=yes, so the failure fixed here only shows under --depwarn=error.
• Merge: needs review

🤖 Generated with Devin CLI 3000.11.3 (model: swe-2-high). Local session, transcript at /home/crackauc/sandbox/goals/qa-hygiene/jobs/gpu-fixes/log.txt on amdci2.