test: stabilize native audio tail parity

[My-Denia]

Description

Stabilizes the Windows native parity harness's final-audio-window analysis without changing either capture backend.

• Extracts PCM tone analysis into a small tested module.
• Uses hysteresis crossings with sub-frame interpolation and 16-cycle local validation.
• Retains the strict 997 +/- 10 Hz requirement.
• Allows at most one AAC frame of below-threshold tail padding and one longest accepted tone cycle after the last valid local window.

Motivation

Valid 997 Hz captures could intermittently fail the final 250 ms parity gate. AAC tail padding and short integer-frame crossing windows made the estimator phase- and quantization-sensitive, even though both helpers produced non-zero AAC packets, decodable frames, and retained audio through pause/resume/stop.

This change fixes the test estimator rather than weakening the audio-quality contract. It does not modify the Rust or legacy C++ capture runtime.

Type of Change

• New Feature
• Bug Fix
• Refactor / Code Cleanup
• Documentation Update
• Other: deterministic parity regression coverage

Related Issue(s)

Release-candidate validation follow-up; no separate issue.

Screenshots / Video

Not applicable. There is no UI change.

Testing

• npx vitest run electron/native/windowsNativeAudioAnalysis.test.mjs - 11/11 passed.
• npm run test - 63 files / 500 tests passed.
• npm run lint - passed.
• npx tsc --noEmit - passed.
• Four retained parity failure artifacts replayed successfully under the strengthened analyzer.
• Three consecutive Windows active-system-audio parity runs passed; each run produced 141 AAC packets and 141 decodable frames from both C++ and Rust helpers.
• Independent execution audit and final Codex review passed with no blocking finding.

Scope

No runtime helper, exporter, version, release-note, tag, or release change is included.

Checklist

• I have performed a self-review of my code.
• Screenshots or videos are not applicable.
• Changelog and issue links are not applicable to this test-only prerequisite.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 80100abfde

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Stop advancing the valid tail marker on mixed-tone windows

When an out-of-band tail starts, the rolling 16-cycle window initially contains mostly valid 997 Hz cycles, so its averaged frequency remains within tolerance and this assignment advances lastValidToneCrossingFrame into the bad tail. For example, a phase-continuous 1008 Hz tail of 1,600 frames reports an approximately 998.45 Hz whole-span frequency and only about 933 trailingWithoutValidToneFrames, passing both parity assertions despite exceeding the 1,073-frame limit; the added 2,000-frame test only passes because the tail is long enough to outlast this transition bias. Require the local window to consist entirely of accepted tone intervals, or otherwise avoid moving the marker past the actual frequency transition.

Useful? React with 👍 / 👎.