LaserLab v0.4.0: true blinded review and dark lab UI

[ProhibitedTV]

Summary

• seal source labels, paths, provenance, calibration results, and control statistics until explicit unblinding
• add integrity-checked one-way unblinding and leak-safe blinded reports, CSV exports, and community bundles
• simplify the desktop workflow to Setup -> Run -> Review -> Compare -> Export
• apply a restrained dark cyberpunk visual system while keeping claims and interpretation conservative
• update documentation, screenshots, package smoke checks, and cross-platform CI

Validation

• Python compilation succeeded
• 11 unit, integration, blinding, bundle, and offscreen GUI tests passed
• git diff --check passed
• Windows LaserLab-v0.4.0-windows-x86_64.zip rebuilt and verified
• LaserLab.exe --smoke passed
• LaserLabCLI.exe --help passed
• sealed artifact and bundle leak scans passed

Closes #12

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7d7becbecc

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]

Hide counts that identify the laser capture

In blinded runs with standard or strict control generation, keeping exact per-capture frame_count in the sealed manifest lets a reviewer infer which capture was labeled laser whenever the captures have different frame counts: the sealed run still exposes total review samples/results, and generated controls are added only for laser frames. For example, with 10- and 20-frame captures and standard controls, the blinded sample total reveals whether 10 or 20 frames received generated controls, defeating source-role sealing before unblind. Consider omitting or coarsening per-capture counts in the sealed manifest, or otherwise avoiding a total that includes generated controls until unblind.

Useful? React with 👍 / 👎.