docs(spec): add custom scan pipeline design (replace ML Kit scanner)

Own the pipeline for full UI control: CameraX custom camera + gallery
→ manual 4-corner editor → setPolyToPoly perspective warp → 5 filters
(original/grayscale/enhance/binarize/brightness-contrast) → existing
Result/save/share/OCR. Pure Android, no native libs. Stage B (manual);
auto edge detection deferred to stage A (OpenCV). Single page first.

Constraint: introduces CAMERA permission (custom camera); ML Kit doc scanner removed, OCR kept.
Confidence: high
Scope-risk: module
Reversibility: moderate
Directive: user reviews spec, then writing-plans.
Tested: design self-review.
Not-tested: no code yet.
Related: 2026-05-30-clearcam-document-scanner-design.md (superseded scanner)

🐙 Autopus <noreply@autopus.co>