style: fix the Runic format check that is red on master

[ChrisRackauckas-Claude]

Please ignore this PR until it has been reviewed by @ChrisRackauckas.

What changed and why

master is red on the Runic format check, so every open PR fails it too. Runic v1.7.0 wants an explicit return on the final expression of the InitializationProblem{iip, specialize} constructor in lib/ModelingToolkitBase/src/problems/initializationproblem.jl. This adds it. That is the whole diff — one line, no version bump (matching #4884, #4824, #4801).

The expression was already the last one in the function body, so adding return is semantically a no-op.

Where it came from

Not a Runic version bump: Runic v1.7.0 was registered 2026-04-22, months before this code landed, and CI has been resolving runic-version: "1" to it the whole time. The regression is a merge with a red check:

• 21f3163..b77c475 ("Make the opts-accepting InitializationProblem method primary") landed on master as part of feat: add SciMLProblemOptions #4893, merged 2026-08-07T17:01:22Z at 79b19d7.
• feat: add SciMLProblemOptions #4893's own Runic / Runic Format Check was FAILURE at merge time: https://github.com/SciML/ModelingToolkit.jl/actions/runs/31200346113/job/92938575650
• The push build of the merge commit is correspondingly red: https://github.com/SciML/ModelingToolkit.jl/actions/runs/31200352099/job/92938595855
• Every PR merged after it inherited the failure (build: bump MTKBase minor, MTK patch already bumped #4900, Route every MTKParameters buffer's cotangent back to its own argument #4905, Enumerate the non-public names ModelingToolkit reaches for #4883 all FAILURE), and every open PR sampled (Generate nlstep_data for the fully implicit DAE path #4912, Register callable parameters as solver nonlinear operators #4909, Avoid piracy in jump dependency traversal #4907, Pin ModelingToolkit's public reexport surface #4881, Scale dynamics constraints as residuals for optimal control #4841, Enforce bounds on observed variables, with a selectable transformation #4840) fails on this one file and nothing else.

Verification

Reproduced CI exactly on a clean origin/master checkout at 731eff6: Runic v1.7.0 (same version CI installs — see the job log, + Runic v1.7.0), invoked the way fredrikekre/runic-action@v1 does, Runic.main(["--check", "--diff", "--verbose", git ls-files -- '*.jl'...]).

Before (on unmodified master), exit code 1:

[ 46/223] Checking `lib/ModelingToolkitBase/src/problems/initializationproblem.jl`  ✖
diff --git a/initializationproblem.jl b/initializationproblem.jl
--- a/initializationproblem.jl
+++ b/initializationproblem.jl
@@ -169,7 +169,7 @@ function InitializationProblem{iip, specialize}(
-    TProb{_iip}(
+    return TProb{_iip}(
...
n files = 223
EXIT_CODE=1

That is the only ✖ in the run — all 223 tracked .jl files were checked and no other file fails.

After this commit, exit code 0:

[ 46/223] Checking `lib/ModelingToolkitBase/src/problems/initializationproblem.jl`  ✔
...
[223/223] Checking `test/substitute_component.jl` ............................ ✔
n files = 223
EXIT_CODE=0

Zero ✖ lines in the whole run.

What I did not verify

I did not run the ModelingToolkit test suite. The change adds return to an expression that was already the function's last, so it cannot change behavior; the test suites, docs build, and downstream jobs are left to CI.