ref(snapshots): Make diffing logic more strict + better handle different sized images when comparing

[NicoHinderling]

odiff only compares the overlapping region of differently-sized images — pixels outside the overlap are silently ignored. This meant dimension changes like width (200→300px) or height (2688→3676px) could report 0% diff.

Fix: pad both images to the max dimensions with transparent pixels before passing to odiff. This normalizes the canvas so odiff always compares same-sized images and correctly counts the non-overlapping area as changed pixels.

Also restores DIFF_THRESHOLD to 0.01 (was temporarily set to 0 for testing).

[NicoHinderling]

• ref(snapshots): Track diff algorithm version on comparison results #112729
• ref(snapshots): Add configurable diff_threshold for snapshot comparison #112728
• ref(snapshots): Make diffing logic more strict + better handle different sized images when comparing #112632 👈 (View in Graphite)
• ref(snapshots): two small UI tweaks #112631
• master

This stack of pull requests is managed by Graphite. Learn more about stacking.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 001b87e. Configure here.}

[NicoHinderling]

@rbro112 FWIW i tested with 0.001 and the results were pretty much the same as 0.01 but with slightly louder (and not necessarily more helpful) feedback on the diff. I think 0.01 is good compromise for now based off my testing

maybe we can try 0 properly once https://github.com/dmtrKovalenko/odiff 's v4.3.3 is actually released, but id rather do that separately

[rbro112]

Would highly recommend a comment documenting why this is .01 as I could see someone coming along and having no context getting confused

[NicoHinderling]

Added a comment explaining the threshold value.