perf(workspace): offload massive grid selection statistics calculation to background compute

[ZhuchkaTriplesix]

Motivation & Problem

When a user selects a massive range in the Data Grid (e.g. 10,000+ cells via Ctrl+A or shift-selection), GridSelectionCalcEngine.compute performs string parsing, number normalization, variance calculation, and distinct set aggregation synchronously on the main thread for the DataGridCalcBar.

Proposed Solution

1. In DataGridCalcBar / GridSelectionCalcEngine:
   • Introduce an adaptive calculation method (computeAdaptive) that runs locally on the main thread when values.length < 5000, and offloads to compute(GridSelectionCalcEngine.compute, values) when selecting larger cell ranges.
2. Add unit tests for adaptive selection stats computation.

Acceptance Criteria

• Large cell selection calculations (>5,000 cells) do not drop frames on the main UI thread.
• Small selections continue to compute synchronously with zero latency.