ci: build and run the unit test suite on Linux/macOS/Windows

[gqf2008]

Summary

The registered unit tests (55+ targets under ENGINE_BUILD_TESTS) were never compiled or run by CI — the flag defaults to OFF and no workflow ever flipped it. This wires the suite into all three CMake-driven workflows:

• Linux (cpu + vulkan matrix): configure with -DENGINE_BUILD_TESTS=ON, build the remaining test targets, run ctest --output-on-failure.
• macOS: same pattern in mac-build.yml.
• Windows: build_windows.ps1 gains a -RunTests switch (forces tests ON, builds everything the -Target pass skipped, runs ctest) so the release presets stay untouched.

Known environment-specific exclusions, each annotated in the workflow:

• audio_utility_api_test on Debug builds: hits a Debug-only ggml softmax NaN assert inside the denoise models (passes in Release, e.g. the Windows job); excluded until the root cause is chased.
• rnnoise/zipenhancer utility tests on the Linux vulkan matrix only: they hard-require a Vulkan device, and runner behavior is non-uniform (backend registers with zero devices on Xeon, sometimes not at all on EPYC). mesa-vulkan-drivers + explicit VK_ICD_FILENAMES are installed for the matrix; the cpu matrix keeps both tests.
• audio_dsp_test istft variant parity: mean tolerance rode at 2.0e-6 and CI measured 2.0074e-06 — loosened to 3e-5 max / 5e-6 mean, still an order of magnitude below one int16 LSB.
• fun_asr_nano_assets_test: compared temp paths as strings; canonical forms now compared (macOS /var symlink, Windows 8.3 short names).
• supertonic_vector_convnext_exp_test: the 5% perf-improvement assertion is noise-bound on shared runners; gated behind SUPERTONIC_ENFORCE_PERF=1, output parity stays unconditional.

Also fixes an istft tolerance edge and a Vulkan ICD discovery issue surfaced by the first run.

Verification

Full matrix green on the fork (Linux cpu + Linux vulkan + macOS + Windows, ~59 tests each):

mac-build: success
Linux Build: success
Nix Build: success
windows-build: success

Local macOS arm64: text_chunking_test, audio_dsp_test, fun_asr_nano_assets_test, supertonic_vector_convnext_exp_test all pass; suite runs in ctest with --output-on-failure and parallelism 4.

[0xShug0]

@gqf2008 Let’s hold off on this PR for now. The current unit tests (and, in fact, most of the test scripts) need a deep cleaning. Many of them should not belong to unittests. I did a test target leakage fix last week, and planned to do more cleanup after the next release.

[gqf2008]

Understood — that matches exactly what the first CI run surfaced: the noisy failures were all utility/perf/parity tests that probably don't belong in unittests to begin with. Marked as draft. Happy to rebase this on top of your test-target cleanup after the next release — or help with the cleanup itself if that's useful.

[0xShug0]

@gqf2008 This PR can be reopend. I moved random tests to a new option ENGINE_BUILD_EXTENDED_TESTS so the current unit tests are relatively clean. Also updated the thresholds in audio_dsp_test to 3.0e-5f and 5.0e-6.

[gqf2008]

Rebased on the post-split main: the unit gate now runs with no per-test exclusions — everything noisy (audio_utility, rnnoise/zipenhancer, audio_dsp istft) moved to EXTENDED_TESTS/MODEL_TESTS by your split, and the thresholds you updated cover the istft drift. PR is now just the CI wiring: ENGINE_BUILD_TESTS=ON + ctest on Linux (cpu+vulkan matrix), macOS, Windows (ps1 -RunTests).

[0xShug0]

@gqf2008 Merged! Thanks!