feat: implement type-safe isomorphic coords helpers

[shadowusr]

What's done?

Implemented a type system for type-safe coords operations.

With these, it becomes much harder to confuse numbers from different spaces or units.

These helpers are broadly used in the upcoming PRs.

For example:

• You can't get intersection of rects with different pixelRatios
• You can't combine two coords, if one is relative to viewport and the other is relative to the whole page
• You can't get a sum of left and top values without explicit type casting

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 97b73c48d6

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]

Handle Size inputs in fractional css→device conversion

fromCssToDevice explicitly accepts Size<"css">, but in the fractional-pixelRatio branch it routes through roundCoords, which only fills fields when top/left exist. For a size object ({ width, height }), this returns an empty object at runtime, so downstream code gets undefined dimensions on non-integer DPR values (for example 1.25/1.5), breaking size-based geometry.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Preserve sign when converting viewport to capture coords

fromViewportToCaptureArea uses getHeight, which applies Math.abs, so it always returns a non-negative offset. Coordinate conversion should preserve direction (it should be the inverse of fromCaptureAreaToViewport), and inputs above the capture-area top should produce negative values; with the current code those values are mirrored to positive and can place/crop geometry on the wrong side.

Useful? React with 👍 / 👎.