test(suites): cut the race run's critical path without dropping a case

[devin-ai-integration]

What and why

make test spends its wall time on a handful of packages whose tests run serially inside the package, so the race run is bounded by the longest package rather than by the core count. Measured on an 8-core box with every corpus present: 10m45s, with the runtime (506s), smt (359s) and workspace model (351s) packages on the critical path.

Two kinds of change, no test or assertion removed:

• Budget-exhaustion cases spin less. TestRuntimeRobustness/nested_flow_that_never_ends (91s) and unguarded_loop_through_a_merge (53s) in internal/exec/runtime ran the default 1,000,000 action steps under the race detector only to reach ErrActionStepLimitExceeded. They now set ctx.maxActionSteps = 1000, as other robustness cases already do; the code path and the typed error are the same.
• Independent subtests run in parallel.
  • TestRefereeCorpus (internal/exec/smt): the corpus cases run under a cases group with t.Parallel(). Each case gets a value copy of the solver (a solve.Solver is configuration; every check starts its own process) and its own refereeTally, merged into the shared tally under a mutex; the group boundary keeps tally.log and the "encoded no case" check after every case has finished. Subtest names gain a cases/ segment; nothing in the tree refers to the old names.
  • TestIncrementalEqualsFresh / TestIncrementalEqualsFreshCorpora (tests/model): each replay owns its workspace and RNG.
  • TestCompiledCalcsAgreeWithInterpreter (internal/frontend/repl): the Go and C targets build into separate TempDirs from separate sessions.
  • TestExprTypeCheckNoStdlibFalsePositives / TestExprTypeCheckPublishedStdlibDefects (internal/workspace/model): each opens its own workspace over a read-only library source.

Per-package race times before → after (measured before the package reorganization; the packages are unchanged in content): runtime 506s → 194s, smt 359s → 200s, model 351s → 235s, repl 232s → 144s, tests/model 220s → 134s. TestRuntimeRobustness 179s → 19s, TestRefereeCorpus 328s → 167s.

How it was verified

• go test -race -pgo=off -count=1 ./internal/exec/runtime ./internal/exec/smt ./internal/workspace/model ./internal/frontend/repl ./tests/model on top of current develop, with the corpora and OPENSYSML_REQUIRE_SMT=1: passes, zero race reports.
• -race -count=2 on tests/model -run TestIncrementalEqualsFresh$ and the model package's -run TestExprTypeCheck.
• gofmt -l . empty, go vet clean on the touched packages.

Checklist

• make test and make lint pass locally (touched packages; full run in CI)
• Tests added or updated for the change
• Documentation extended where it already covers the surface (not applicable)
• Changelog entry added (test-only change)
• baselines regenerated (no gate count moved)
• No internal work-item labels in the body, docs, or changelog

Link to Devin session: https://nasa-jpl-demo.devinenterprise.com/sessions/9c85d2f2974848dbb25a944ea7212e1e
Open in Devin Desktop: https://nasa-jpl-demo.devinenterprise.com/desktop/session/9c85d2f2974848dbb25a944ea7212e1e?variant=devin
Requested by: @HuiJun

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.