fix(sqlite): columns with the same name collapse into one in query results #1144

Problem

A SQLite result with two columns of the same name shows only one of them, and its values may come from the other column. No error, the user just sees wrong data.

Reproduced on the demo database (via MCP, same path as the SQL editor):

SELECT o.id, u.id FROM orders o JOIN users u ON u.id = o.user_id LIMIT 1
-- columns: ["id"], one value — expected two columns
SELECT o.*, u.* FROM orders o LEFT JOIN users u ON u.id = o.user_id ORDER BY o.id LIMIT 1 OFFSET 11
-- shows id = 2 (users.id) and users.created_at for order 12

Cause

SqliteConnection.execute returns List<Map<String, Object?>> from sqflite's rawQuery; SqliteSqlExecutionDelegate.executeQuery builds the column list from results.first.keys and reads cells by name. Equal names map to one key. Same class of bug as the ERD catalog fixed in #1139.

Affected

• SQL editor results for SQLite (any JOIN / * with repeated names, SELECT 1, 1, expressions without aliases that print the same).
• MCP run_query / sample_rows on SQLite connections (epic(mcp): Querya Desktop as a read-only MCP server #1133).
• Anything else that reads SQLite rows by name (table browser, export of such results).

Fix ideas

• Read results positionally: column list + rows as lists. Options: sqflite's QueryResultSet keeps columns / rows internally; or run read queries through package:sqlite3 (ResultSet.columnNames + rows), which sqflite_common_ffi already depends on.
• Keep display names as SQLite reports them (id, id); the grid already uses positions.

Acceptance

• SELECT o.id, u.id ... returns two columns with the right values in the SQL editor and over MCP.
• Regression tests on a real SQLite file: duplicate names, SELECT 1, 1, o.*, u.*.
• No change for results with unique names.