replay: scrubbing a tape keeps the glass's edge

[foxnne]

What changes

Scrubbing a tape no longer drops the glass's edge on the transport bar and the corner controls.

Cause. A seek replays to its moment with the app's motion skipped (Stage.fastForward → Demo.fast), and Editor.tick did that by publishing motion as off, the same as a system asking for reduced motion. But glass's edge (its lens, rim light and bend) is an amount the motion level grants (motion.liquid, read by BlurBackdrop.queuePane and DropZones). So every frame drawn during a seek drew all glass flat. It showed most across the tour's Materials chapter, where seeks take several displayed frames to land.

Fix. core.motion.publish takes skipping separately from the system's preference. Animations still land at once, as at off, while liquid stays at the person's level. liquid is published under its own key, so a plugin built before this change still reads the level as before (with the old behaviour) and no shared struct changes shape.

The coordinator also suspected the material. It is already part of the replay state: snapshots and keyframes hold window_material, and the snapshot fingerprint covers it. The test now also checks that.

SDK impact

• None
• Core-only or additive: reaches plugins at the next SDK release
• Fingerprint moved: recorded in sdk/src/version.zig, sdk_version untouched, PR labelled sdk
• SDK release: bumps sdk_version, lists the sdk PRs since the last sdk-v* tag

Verified

• macOS: the headless tour test (plays the tour) now scrubs across the Materials chapter both ways with a seek budget of 0, so every displayed frame is one of the seek's. On every one of those frames, motion is skipped and motion.liquid() equals its value in straight playback. Every landing replays with no fingerprint mismatches. With the old publish put back, the test fails (expected 1, found 0). Gates run locally: check-fmt, build, test, test-integration, check-web. I did not take a native screenshot.
• Windows:
• Linux: CI
• Web: check-web only

Follow-ups

• A plugin built against an older core still computes liquid from the level, so its own glass (pixi's, for one) loses its edge during a seek until it is repinned at the next SDK release.
• Seek frames now draw glass with its edge, as every other frame does, instead of flat. On native, where silent runs draw everything anyway, that costs the same as a normal frame. I did not measure it separately.

🤖 Generated with Claude Code