[FabioHAraujo]

Closes #10
Closes #11
Closes #12
Closes #13
Closes #14

Problem and resulting behavior

There was no way to move data in or out of LowQuery beyond copying grid rows. This adds export and import that stay simple: few choices, sensible defaults, everything detected that can be.

• Export… (tables, views, other objects, schemas): format (SQL, CSV, CSV for Excel, JSON) and, for SQL, structure and data / structure / data. Then where to save. A schema becomes one .sql file, or one CSV/JSON per table.
• Import data… (table) and Import file as a new table… (schema): CSV or JSON with encoding, delimiter and header detected, columns matched by name, spreadsheet-style numbers and dates understood, one transaction. Options: delimiter, header row, replace the rows.
• Run SQL file… (connection, schema): runs a script with progress, stops at the first error unless unticked, lists failures.

Backend: Session::stream reads rows from each driver's cursor in batches (values in full, not cut to the grid preview) into a channel; transfer/ writes SQL (structure from object_ddl, tables in foreign key order, exact literals per engine, identity handling, generated columns skipped), CSV and JSON, and imports with the same literal writer.

Validation

• transfer/round_trip.rs on every database in CI: a table with quotes, line breaks, backslashes, accents, binary, JSON, dates, identity and generated columns is exported and loaded back from the SQL script (after dropping it), CSV, CSV for Excel and JSON (after deleting its rows); rows must match exactly, and the identity generator must accept a new row afterwards.
• Unit tests for literals, CSV parsing, encodings, number/date conversion, type inference and foreign key order; bun test, svelte-check, clippy, fmt.
• Dialogs checked in the browser with the development mock, and the commands in a debug build of the app against a real SQLite file (SQL, CSV for Excel and JSON exports, schema export to a folder, CSV import back with quotes, accents and NULL intact).

Additional context

Includes the Oracle ORA-01002 fix from #18 (already on main) for the 0.5.2 release. Oracle binary values over 2000 bytes cannot be written as a SQL literal yet; CSV/JSON export them fine.