Fix KllItemsSketch sorted-view corruption after query-before-serialize

[jaideeppyne]

Fixes #756.

Problem

Querying a heap KllItemsSketch before serializing it makes heapify / wrap return a corrupted sorted view (wrong quantiles, duplicated min/max, wrong retained count in the view).

Root cause

KllItemsSketch.CreateSortedView.getSV() sorts level-0 and then records that it did:

final T[] srcQuantiles = getTotalItemsArray();
...
if (!isLevelZeroSorted()) {
  Arrays.sort(srcQuantiles, srcLevelsArr[0], srcLevelsArr[1], comparator);
  if (!hasMemorySegment()) { setLevelZeroSorted(true); }
}

For the heap Items variant, KllHeapItemsSketch.getTotalItemsArray() returns a defensive copy, so the sort lands on the copy while setLevelZeroSorted(true) is set on the live sketch. Serialization writes that flag; heapify / wrap trust it and skip sorting.

The Doubles path is fine because KllHeapDoublesSketch.getDoubleItemsArray() returns the live array, which is what makes the comment at KllDoublesSketch CreateSortedView true.

Fix

Minimal change: do not call setLevelZeroSorted(true) in Items CreateSortedView. Only a copy was sorted, so the live level-0 remains unsorted and the serialized flag must stay false. An alternative would be returning the live array from getTotalItemsArray() to match Doubles; this PR prefers not claiming the live sketch was sorted when it was not.

Tests

Added regression tests in KllItemsSketchSerDeTest:

• query then heapify — sorted view quantiles/weights and probed ranks match
• query then wrap — same
• query after enough updates to force compaction (k=8, n=8) — heapify and wrap match

Also asserts isLevelZeroSorted() stays false after the query on the heap sketch.

AI disclosure

I used Cursor / Grok to help draft the patch and open this PR via the GitHub API (no local clone). I verified the root cause against the sources cited in #756 before editing.

[leerho]

Thank you for finding and fixing this!