fix: reference Symbolics-owned names through Symbolics in tests

SciML#4989 trimmed `hessian_sparsity`, `MultithreadedForm`, `SerialForm`, `CTarget`
and `derivative` from ModelingToolkitBase's Symbolics import list, as
ExplicitImports reported them unused in `src`. They were still reached as
`ModelingToolkitBase.<name>` from tests, and none is exported by Symbolics, so
the re-export does not cover them. That broke InterfaceII and Extended.

Reference them through Symbolics, which owns them.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>

Author: ChrisRackauckas

lib/ModelingToolkitBase/test/bigsystem.jl

@@ -57,7 +57,7 @@ f(du, u, nothing, 0.0)
 multithreadedf = eval(
     ModelingToolkitBase.build_function(
         du, u, fillzeros = true,
-        parallel = ModelingToolkitBase.MultithreadedForm()
+        parallel = Symbolics.MultithreadedForm()
     )[2]
 )
 
@@ -76,8 +76,8 @@ end
 
 #=
 jac = sparse(ModelingToolkit.jacobian(vec(du),vec(u)))
-fjac = eval(ModelingToolkit.build_function(jac,u,parallel=ModelingToolkit.SerialForm())[2])
-multithreadedfjac = eval(ModelingToolkit.build_function(jac,u,parallel=ModelingToolkit.MultithreadedForm())[2])
+fjac = eval(ModelingToolkit.build_function(jac,u,parallel=Symbolics.SerialForm())[2])
+multithreadedfjac = eval(ModelingToolkit.build_function(jac,u,parallel=Symbolics.MultithreadedForm())[2])
 
 u = rand(N,N,3)
 J = similar(jac,Float64)
@@ -96,7 +96,7 @@ maximum(J2 .- Array(J)) < 1e-5
 jac = ModelingToolkitBase.sparsejacobian(vec(du), vec(u))
 serialjac = eval(ModelingToolkitBase.build_function(vec(jac), u)[2])
 #multithreadedjac = eval(ModelingToolkit.build_function(vec(jac), u,
-#    parallel = ModelingToolkit.MultithreadedForm())[2])
+#    parallel = Symbolics.MultithreadedForm())[2])
 
 MyA = zeros(N, N)
 AMx = zeros(N, N)

lib/ModelingToolkitBase/test/ccompile.jl

@@ -1,4 +1,5 @@
 using ModelingToolkitBase, Test
+using Symbolics: CTarget
 using ModelingToolkitBase: t_nounits as t, D_nounits as D
 
 @parameters a
@@ -9,7 +10,7 @@ eqs = [
 ]
 f = build_function(
     [x.rhs for x in eqs], [x, y], [a], t, expression = Val{false},
-    target = ModelingToolkitBase.CTarget()
+    target = CTarget()
 )
 f2 = eval(build_function([x.rhs for x in eqs], [x, y], [a], t)[2])
 du = rand(2);

lib/ModelingToolkitBase/test/nonlinearsystem.jl

@@ -1,4 +1,5 @@
 using ModelingToolkitBase, StaticArrays, LinearAlgebra
+using Symbolics: hessian_sparsity
 using DiffEqBase, SparseArrays
 using Test
 using NonlinearSolve
@@ -92,9 +93,9 @@ jac = calculate_jacobian(ns)
 jac = generate_jacobian(ns)
 
 sH = calculate_hessian(ns)
-@test getfield.(ModelingToolkitBase.hessian_sparsity(ns), :colptr) ==
+@test getfield.(hessian_sparsity(ns), :colptr) ==
     getfield.(sparse.(sH), :colptr)
-@test getfield.(ModelingToolkitBase.hessian_sparsity(ns), :rowval) ==
+@test getfield.(hessian_sparsity(ns), :rowval) ==
     getfield.(sparse.(sH), :rowval)
 
 prob = NonlinearProblem(ns, [x => 1.0, y => 1.0, z => 1.0, σ => 1.0, ρ => 1.0, β => 1.0])

test/reduction.jl

@@ -267,7 +267,7 @@ sys = mtkcompile(sys0)
 eq = equations(tearing_substitution(sys))[1]
 vv = only(unknowns(sys))
 @test isequal(eq.lhs, D(vv))
-dvv = ModelingToolkit.value(ModelingToolkit.derivative(eq.rhs, vv))
+dvv = ModelingToolkit.value(Symbolics.derivative(eq.rhs, vv))
 @test dvv ≈ -60
 
 # Don't reduce inputs