Fix ModelingToolkitBase Aqua unbound arguments

[ChrisRackauckas-Claude]

Summary

• remove unused static parameters from the parameter-buffer signatures reported by Aqua
• replace the two unbound rm_union methods with public typeintersect while preserving numeric array element types
• add regression coverage for Union{Nothing, BigFloat} and Union{Nothing, Float32} arrays

This is a focused part of #4670. It does not suppress or ignore Aqua findings.

Local verification

• GROUP=SymbolicIndexingInterface: 64/64 SymbolicIndexingInterface, 1804/1804 SciML problem inputs, 118/118 MTKParameters; exit 0
• direct variable_utils.jl: parse-variable testsets 54/54 and 54/54, isinitial 7/7, At 13/13; exit 0
• focused Aqua.test_unbound_args(ModelingToolkitBase): exit 0
• official GROUP=QA: JET 54/54; Aqua 7 pass / 4 fail, with unbound-argument checks passing. The full QA command exits 1 because this independent branch intentionally leaves the ambiguity, stale-dependency, missing-compat, and piracy classes for separate fixes.
• Runic.jl 1.7.0 over the full repository: exit 0

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

[ChrisRackauckas-Claude]

Rebased onto current upstream aa4afff870 (including merged #4737 metadata fixes). Focused verification after the rebase, Julia 1.12.6:

Aqua.test_unbound_args(ModelingToolkitBase) exited 0 and printed unbound_aqua_passed.

git diff --check upstream/master...HEAD also exits 0. New head: d31e9905ab. The update used an exact force-with-lease after fetching and inspecting remote head 7f5f7094a6.

[ChrisRackauckas-Claude]

Fresh rebased CI confirms this PR fixes its intended Aqua class in the full QA suite: Unbound type parameters passes. The QA job is red only for the independently tracked method-ambiguity and piracy classes documented in #4670 (Aqua summary: 9 passed, 2 failed). Job: https://github.com/SciML/ModelingToolkit.jl/actions/runs/29254919662/job/86834616664. The other red check is the exact clean-main piston/PleaseImportDynamicQuantities downstream fixture currently under its separate MTKSL/MTK reproduce-bisect-fix audit: https://github.com/SciML/ModelingToolkit.jl/actions/runs/29254919009/job/86832559956.

[ChrisRackauckas-Claude]

Additional rebased-CI classification:

• Downgrade/InterfaceI exactly selects local MTKB 1.51.1, Symbolics 7.30.0, and SymbolicUtils 4.37.0, then hits the separate clean-main UndefVarError: CodegenFunctionOptions not defined at codegen_utils.jl:636; that dependency-boundary audit is active. Job: https://github.com/SciML/ModelingToolkit.jl/actions/runs/29254919457/job/86832560838.
• The LTS benchmark computation completed; only create-or-update-comment failed with HTTP 403 Resource not accessible by integration, the known workflow-token permission issue. Job: https://github.com/SciML/ModelingToolkit.jl/actions/runs/29254918887/job/86832559442.
• The current benchmark job was marked cancelled while still running, with no benchmark assertion/package error. Job: https://github.com/SciML/ModelingToolkit.jl/actions/runs/29254918887/job/86832559474.
• The root QA job again has no uploaded log blob (BlobNotFound), so it reports no actionable package failure. Job: https://github.com/SciML/ModelingToolkit.jl/actions/runs/29254919728/job/86833241565.

None is caused by the unbound-signature patch; the full MTKB QA lane confirms that intended Aqua class passes.

[ChrisRackauckas-Claude]

The remaining newly completed sublibrary reds are existing clean-main clusters:

• MTKB downgrade is another CodegenFunctionOptions precompile failure under the Symbolics floor; it belongs to the active public-API prerequisite audit.
• InterfaceII on both Julia LTS and pre passes 7,237 tests, then reports the exact same two Float64(::ForwardDiff.Dual) BVP errors at bvproblem.jl:14 and :308. Those were reproduced on clean MTK master and fully validated against the candidate in closed BoundaryValueDiffEq build_function is allocating for StaticArrays #559: Use finite-difference AD for BVP default nonlinear solvers BoundaryValueDiffEq.jl#559 (comment). build_function is allocating for StaticArrays #559 was closed unmerged by the maintainer and will not be reopened or duplicated here.

Jobs: https://github.com/SciML/ModelingToolkit.jl/actions/runs/29254919442/job/86832704083, https://github.com/SciML/ModelingToolkit.jl/actions/runs/29254919662/job/86834616845, https://github.com/SciML/ModelingToolkit.jl/actions/runs/29254919662/job/86834616620.