fix(segmentation): replace unbounded render queue with coalescing Set

[pedrokohler]

Context

SegmentationRenderingEngine._pendingRenderQueue is a string[][] that grows
without bound. Every scroll event on a StackViewport with segmentations appends a
new viewport-ID array to the queue. The queue is drained one entry per animation
frame via shift(), but entries arrive far faster than they are consumed.

After 310 scroll events on a viewport with 13 labelmap segmentations, the queue
held 4,077 entries — each an array of 13+ viewport IDs with associated
closures and promise chains — consuming significant memory and CPU.

Changes & Results

Replace _pendingRenderQueue: string[][] = [] with
_pendingRender: Set<string> = new Set().

• Duplicate viewport IDs are coalesced into a single entry in the Set.
• When the animation frame fires, all pending IDs are flushed in one batch and the
  Set is cleared.
• Queue size is now bounded to O(viewports) instead of O(scroll events).

Before: queue grows linearly with scroll events (4,077 entries after 310 scrolls).
After: pending set stays bounded at the number of unique viewport IDs (e.g. 13).

Testing

1. Open a StackViewport with 10+ labelmap segmentations.
2. Scroll rapidly through 300+ slices.
3. Observe stable memory — no unbounded growth in the render queue.

Checklist

PR

• My Pull Request title is descriptive, accurate and follows the
  semantic-release format and guidelines.

Code

• My code has been well-documented (function documentation, inline comments,
  etc.)

Public Documentation Updates

• The documentation page has been updated as necessary for any public API
  additions or removals.

Made with Cursor