perf(sqlite): eliminate redundant double string conversion pass in SqliteSqlExecutionDelegate #1354

Summary

In lib/features/sqlite/sqlite_sql_workspace.dart (lines 108–114):

final rawRows = results.take(limitCount).map((row) {
  return cols
      .map((col) => sqliteResultCellToDisplayString(row[col]))
      .toList();
}).toList();

final outRows = await convertResultRowsToStringsAdaptive(rawRows);

1. rawRows is already converted into a List<List<String>> using sqliteResultCellToDisplayString.
2. Immediately on line 114, rawRows is passed into convertResultRowsToStringsAdaptive(rawRows).

As explicitly documented in postgres_result_utils.dart (lines 39–42):

"Rows are already strings after this, so callers must not run them through convertResultRowsToStringsAdaptive again: that would walk (and, from 1000 rows, copy to an isolate) the whole matrix a second time without being able to recover any type information."

For SQLite queries returning $\ge 1000$ rows, convertResultRowsToStringsAdaptive copies the entire 2D string matrix to a worker isolate and back, allocating a duplicate matrix of identical strings and doubling the execution CPU cost.

Proposed Solution

1. Pass raw objects directly to convertResultRowsToStringsAdaptive with a single-pass conversion, or:
2. Use rawRows directly without calling convertResultRowsToStringsAdaptive a second time (or intern strings during the first pass).

Acceptance Criteria

• SQLite query execution performs only a single pass over result rows.
• Eliminates isolate serialization/deserialization overhead for SQLite result sets $\ge 1000$ rows.
• Displayed grid cells and cell editing remain identical.