Fix remaining CI failures: downstream MATLAB ordering & SymPy PyCall setup

[ChrisRackauckas-Claude]

Summary

Fixes the remaining failing CI jobs on master (other than the Core reference tests handled in #1878).

1. test (Downstream, 1) — MATLAB target regression test

test/downstream/ParameterizedFunctions_MATLAB.jl compared the generated MATLAB string byte-for-byte. The order of operands within a commutative * follows SymbolicUtils' argument sort, which is benign but version-dependent: the expected string has u(1) * u(2) while current stable Julia (1.12) emits u(2) * u(1). (The test already looked up parameter indices to stay stable across parameter reorderings — this extends the same robustness to factor ordering.)

Fix: compare the two RHS rows modulo commutative +/* operand ordering (mirrors the build_targets tests from #1878).

Verified: passes locally 2/2 on Julia 1.12, and the Downstream CI job is green on this PR.

2. test (SymPy, 1) — PyCall/sympy precompilation

The SymPy group failed with ModuleNotFoundError: No module named 'sympy'. PyCall defaulted to the runner's system Python (/usr/bin/python3), which has no sympy, so SymbolicsSymPyExt failed to precompile. sympy.jl already tried ENV["PYTHON"]="" + Pkg.build("PyCall") to switch to the Conda.jl Python, but it ran after Pkg.instantiate() had already precompiled the extension against the system Python — too late, since PyCall's Python is fixed at load time.

Fix: set ENV["PYTHON"]="" before instantiation in activate_sympy_env (CI only), so PyCall builds against the Conda Python and SymPy's pyimport_conda installs sympy. CI logs confirm this now works (Installing sympy via the Conda sympy package... sympy-1.14.0, extension precompiles).

3. SymPy solver bug uncovered by fix #2 (sympy.jl:97)

Fixing the precompilation exposed a latent correctness bug that had been masked by the precompile failure. integrating_factor_solve guarded against SymPy returning an unevaluated Integral(...) with get_variables(solution, variable(:Integral)) — but the unevaluated integral comes back as an operation, not a variable, so the guard never fired and sympy_simplify mangled it into a wrong closed form.

Concretely symbolic_solve_ode(Dt(x) + sqrt(t)*x ~ sin(4t)*x^3, x, t) returned 1/sqrt(-2sin(4t) + C₁*exp((4//3)*t^(3//2))) (which does not satisfy the ODE — I verified the residual is non-zero) instead of nothing.

Fix: detect the unevaluated integral on the printed form of the solution and return nothing, so such ODEs correctly report no closed-form solution.

Verified: test/sympy.jl passes locally 16 passed / 1 broken (pre-existing) / 0 failed on Julia 1.12 with sympy 1.14.0, against a dev build of this branch (the solvable Bernoulli cases still solve correctly — no regression).

Not addressed here

• benchmark: this job's benchmarks run fine; only the final PR-comment step fails with 403 Resource not accessible by integration, because GITHUB_TOKEN is read-only for fork PRs (a GitHub security restriction, independent of the permissions: block). It does not run on master and is not a code/test failure, so I left the workflow unchanged rather than risk masking real benchmark regressions. It will show red on this PR for the same fork-token reason.

---

Please ignore until reviewed by @ChrisRackauckas.

🤖 Generated with Claude Code

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 24.87%. Comparing base (c736678) to head (f60de33).
⚠️ Report is 107 commits behind head on master.
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1880      +/-   ##
==========================================
- Coverage   26.12%   24.87%   -1.25%     
==========================================
  Files          56       57       +1     
  Lines        5356     5589     +233     
==========================================
- Hits         1399     1390       -9     
- Misses       3957     4199     +242     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.