Import public API from its owning module, not a re-exporter

[ChrisRackauckas-Claude]

Please ignore until reviewed by @ChrisRackauckas.

Problem

On unmodified master (32a955a), the QA lane's ExplicitImports checks fail. Two of them are about provenance: ModelingToolkit reaches ten bindings through a package that merely forwards them, rather than through the module that declares them.

########## explicit_imports_via_owners
Module `ModelingToolkit` has explicit imports of names from modules other than their owner:
- `@add_kwonly` has owner `SciMLBase` but it was imported from `DiffEqBase`
- `Operator` has owner `SymbolicUtils` but it was imported from `Symbolics`
- `Term` has owner `SymbolicUtils` but it was imported from `Symbolics`
- `_isone` has owner `SymbolicUtils` but it was imported from `Symbolics`
- `_iszero` has owner `SymbolicUtils` but it was imported from `Symbolics`
- `getname` has owner `SymbolicIndexingInterface` but it was imported from `Symbolics`
- `maketerm` has owner `TermInterface` but it was imported from `SymbolicUtils`
- `metadata` has owner `TermInterface` but it was imported from `SymbolicUtils`

########## qualified_accesses_via_owners
Module `ModelingToolkit` has qualified accesses to names via modules other than their owner:
- `NullParameters` has owner `SciMLBase` but it was accessed from `DiffEqBase`
- `value` has owner `SciMLBase` but it was accessed from `DiffEqBase`

Fixing those exposed four more in the ModelingToolkit.StructuralTransformations submodule (ExplicitImports reports the first failing module only): topological_sort (owner Graphs), value and var_from_nested_derivative (owner Symbolics), and schedule (owner Base), all imported from ModelingToolkitBase.

Change

Each name is imported from its owner. schedule had no use site in StructuralTransformations (only sys.schedule and get_schedule appear), so it is dropped rather than re-pointed. TermInterface becomes a direct dependency; it was already in the manifest via SymbolicUtils.

Verification

All 13 relocated bindings are === to the ones previously used, so this is a pure import-site change:

@add_kwonly                   true
Operator                      true
Term                          true
_iszero                       true
_isone                        true
getname                       true
maketerm                      true
metadata                      true
NullParameters                true
value                         true
topological_sort              true
MTKBase.value                 true
var_from_nested_derivative    true
ALL IDENTICAL = true

Both checks now pass (run against this branch on Julia 1.11):

########## explicit_imports_via_owners
OK

########## qualified_accesses_via_owners
OK

Six of the moved names are public at their owner but not at the re-exporter, so the two remaining *_are_public checks also improve, from 57 to 53 explicit-import findings and from 91 to 89 qualified-access findings. Those two checks are not closed by this PR.

Functional checks run locally against this branch:

• test/semilinearodeproblem.jl (the file containing the DiffEqBase.value -> SciMLBase.value change): Test Summary: Astrochem model | 61 Pass 6 Broken 67 Total 5m57.7s
• Ad-hoc smoke: Lorenz mtkcompile + ODEProblem + solve(Rodas5P()), index-3 pendulum mtkcompile (exercises StructuralTransformations), and linearization_function/linearize all pass.

🤖 Generated with Claude Code

https://claude.ai/code/session_019ywCW8vbGoc9X3dUbmXyme