[MOL-21512][TIEN] refactor compression flow for image-upload

[tientranngoc-alt]

Type of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing apis or functionality to change)
• Documentation (change to documentation, comments or API descriptions)
• Tests (improvements to unit tests or E2E tests)
• Other (technical improvements, refactoring, or changes that don't fall into the above categories)

Description of changes

• Link to ticket

• Root cause: compressImage called convertBlob() which decoded and re-encoded the image at full resolution
  (e.g. 2634×4680) before resizing => wasting 2–4 seconds per upload on a redundant encode cycle.

• Fix

  • eliminate redundant full-res re-encode: Replace convertBlob() (decode + full-res encode) with convertHeicToBlob() (HEIC conversion only) + blobToImage() (decode once). The image is drawn directly onto a resized canvas, skipping the wasteful intermediate step.

  • binary search compression with canvas reuse: Replace the old linear quality-reduction loop (up to 10 decode/encode iterations) with a binary search algorithm (4–6 encode-only iterations). The pre-drawn canvas is passed directly to compressToTargetSize(), eliminating repeated decoding.

  • early exit: Skip compressToTargetSize() entirely if the resized blob already fits within maxSizeInKb.

Checklist

• Changes follow the project guidelines in CONTRIBUTING.md
• Looks good on mobile and tablet
• Updated documentation
• Added/updated unit tests

Screenshots

Screen.Recording.2026-07-24.at.14.13.36.mov