Allow Symbolics facade reexports in MTKBase QA

[ChrisRackauckas-Claude]

Ignore this PR until it has been reviewed by @ChrisRackauckas.

What changed and why

Symbolics 7.38 replaced its blanket SymbolicUtils reexport with explicit exports. That ownership correction made ModelingToolkitBase's strict reexport audit newly identify Unknown, scalarize, shape, and unwrap, even though all four are already part of the intentional ModelingToolkitBase/Symbolics facade API.

This adds those existing SymbolicUtils-owned facade names to REEXPORTED_API; it does not widen the user-facing API. This is the ModelingToolkitBase counterpart to #5039.

Verification

Failing before the change on clean upstream/master at f9131984d97460057c62830b60790380ee5fc2f1 with Julia 1.12.7 and Symbolics 7.39.0:

$ GROUP=QA julia +1.12 --startup-file=no --project=. -e 'using Pkg; Pkg.test()'
No unapproved public reexports: Test Failed
  Evaluated: isempty([:Unknown, :scalarize, :shape, :unwrap])
Test Summary: | Pass  Fail  Total
Aqua Tests    |   19     2     21

The other failure was the pre-existing package-wide JET result (263 possible errors), tracked at #4958. The exact JET lane was already red with the same 263 reports in its first CI run after introduction by 92c27bb:
https://github.com/SciML/ModelingToolkit.jl/actions/runs/31478008118/job/93737918032

Focused reexport audit after the change:

UNAPPROVED_REEXPORTS=Symbol[]

Full post-change QA:

$ GROUP=QA julia +1.12 --startup-file=no --project=. -e 'using Pkg; Pkg.test()'
═════ 263 possible errors found ═════
Test Summary:                      | Pass  Fail  Total
Aqua Tests                         |   20     1     21
  No unapproved public reexports   |    1            1

Formatting and spelling:

$ julia +1.12 --project=@runic -e 'using Runic; exit(Runic.main(ARGS))' -- --check lib/ModelingToolkitBase/test/qa/aqua.jl
$ typos lib/ModelingToolkitBase/test/qa/aqua.jl
$ git diff --check

All three completed with exit code 0 and no output.

Not verified

• Full QA remains blocked by the pre-existing JET generated-guard reports; no test was skipped, disabled, or silenced. The upstream reductions are Package analysis exhausts memory; generated short-circuit guards produce false-positive UndefVarError reports aviatesk/JET.jl#858 and [inference] retain undef phi for short-circuit-bound local in nested && guard JuliaLang/julia#62745.
• Docs were not built because this changes only the QA allowlist and neither source, docstrings, public API, nor docs/.
• GPU, downstream, Julia pre-release, and allowed-to-fail jobs were not run locally.
• No dependency or compatibility entry changed.

Links

• Symbolics 7.38 release: https://github.com/JuliaSymbolics/Symbolics.jl/releases/tag/v7.38.0
• Parallel root-package fix: Allow intentional Symbolics facade reexports in QA #5039
• MTK JET tracking issue and history evidence: QA: JET package analysis exceeds the 60-minute root QA budget #4958 (comment)
• Upstream JET issue: Package analysis exhausts memory; generated short-circuit guards produce false-positive UndefVarError reports aviatesk/JET.jl#858
• Upstream Julia issue: [inference] retain undef phi for short-circuit-bound local in nested && guard JuliaLang/julia#62745

🤖 Generated with Codex (harness version: unknown; model: gpt-5; session: local session ID 01a04feb-43da-7ce1-99e4-a68438c69833).

[ChrisRackauckas-Claude]

Implementation scratchpad:

1. Reproduced clean-master QA: 19 pass / 2 fail, with the four SymbolicUtils reexports plus the pre-existing 263-report JET failure.
2. Traced the reexport boundary to Symbolics 7.38.0 and the JET boundary to MTKBase strict-QA introduction in commit 92c27bb; its first CI run already had the same 263 reports: https://github.com/SciML/ModelingToolkit.jl/actions/runs/31478008118/job/93737918032.
3. Added only the four existing facade names to the MTKBase allowlist.
4. Post-change full QA: 20 pass / 1 fail; the reexport check passes, and only the tracked JET failure remains. Runic, typos, and diff checks pass.
5. Awaiting CI; PR remains draft and should be ignored until reviewed by @ChrisRackauckas.

JET investigation update: #4958 (comment)

[ChrisRackauckas-Claude]

CI triage update:

The Runic workflow failed on three files unchanged by this PR:

• lib/ModelingToolkitBase/ext/MTKBifurcationKitExt.jl
• lib/ModelingToolkitBase/src/systems/parameter_buffer.jl
• lib/ModelingToolkitBase/test/accessor_functions.jl

The changed file, lib/ModelingToolkitBase/test/qa/aqua.jl, passed the CI Runic check. This branch differs from upstream/master only in that QA file. The failed workflow is https://github.com/SciML/ModelingToolkit.jl/actions/runs/33300210480/job/99226805030.

Per repository policy, a separate clean-master formatting investigation is in progress; this PR will not mix unrelated formatting changes into the reexport metadata fix.

[ChrisRackauckas-Claude]

Runic follow-up: this is the known Runic 1.10 formatter drift already isolated in draft #5046. Clean upstream/master at f9131984d9 fails Runic 1.10.0 on exactly the same three unchanged files; Runic 1.7/1.8/1.9 pass. The changed lib/ModelingToolkitBase/test/qa/aqua.jl passes both local Runic and the CI Runic check. PR 5047 therefore depends on the separate mechanical PR 5046 for a green repository-wide formatter job.

[ChrisRackauckas-Claude]

CI triage is complete.

• ModelingToolkitBase QA reached Aqua Tests | 20 pass, 1 fail, 21 total; No unapproved public reexports | 1 pass. The sole failure is the clean-master package-wide JET result, 263 possible errors found: https://github.com/SciML/ModelingToolkit.jl/actions/runs/33300210564/job/99229948752
• Root ModelingToolkit QA reached 39 pass, 1 fail; its sole failure is the same four root-package reexports. That parallel scope is covered by draft PR Allow intentional Symbolics facade reexports in QA #5039. Job: https://github.com/SciML/ModelingToolkit.jl/actions/runs/33300210622/job/99228462845
• Downgraded ModelingToolkitBase QA also passes this PRs reexport assertion, but fails pre-existing downgrade-environment checks for stale dependencies, circular-precompile/persistent tasks, and an implicit UnPack import. The clean-master-only formatting PR reproduces the same failures: https://github.com/SciML/ModelingToolkit.jl/actions/runs/33298117906/job/99221298490. This PRs job: https://github.com/SciML/ModelingToolkit.jl/actions/runs/33300210594/job/99228275703
• Runic remains dependent on the separate clean-master Runic 1.10 drift fix in draft PR Format files with Runic 1.10 #5046; this PRs changed aqua.jl file passes Runic 1.10.

No failure unique to this patch was found.