perf(calc): optimize GridSelectionCalcEngine median calculation with QuickSelect for large selections

[ZhuchkaTriplesix]

Problem

In lib/features/main_screen/grid_selection_calc_engine.dart, the compute method calculates the median of a multi-cell selection by copying and performing a full (K \log K)$ sort on the entire numericList:

numericList.sort();
final mid = numericCount ~/ 2;
if (numericCount.isOdd) {
  median = numericList[mid];
} else {
  median = (numericList[mid - 1] + numericList[mid]) / 2.0;
}

When selecting tens of thousands of cells across entire columns, full sorting creates unnecessary CPU overhead when only the 50th percentile (median) value is required.

Proposed Solution

• Introduce a linear-time QuickSelect ((K)$ average time) algorithm for median calculation when the numeric list exceeds a threshold (e.g. > 500$).
• Keep standard fast sort for small selections ( \le 500$).
• Add unit tests in grid_selection_calc_engine_test.dart validating median correctness for odd/even selections and benchmark performance.