fix(core): build chunks by appending so oversized events don't stall …

[abueide]

…the queue

chunk() assigned batches by index (chunks[++currentChunk] = [item]). When the first item alone was >= maxKB, currentChunk went 0 -> 1 and index 0 was never written, producing a sparse array. Since 2.23.0 the upload path iterates batch results with for...of, which does not skip holes, so aggregateErrors read .status off undefined and every flush threw. The throw escapes sendEvents after the upload but before processUploadResults, so nothing is ever dequeued and the queue only recovers when pruneExpiredEvents discards the events 12 hours later.

rollingKBSize was also never reset when a new chunk started, so once the cumulative size crossed maxKB every remaining item became its own batch and the count limit became unreachable.

Build the chunks by appending and reset the accumulator per chunk. This removes the hole, restores the count limit, and keeps an item that alone exceeds maxKB isolated in its own batch so the server rejects it and the queue continues to drain.