Change the database schema

SCHEMA_VERSION lives in src-tauri/src/db/migrations.rs. src-tauri/src/db/mod.rs owns the connection pool and init only. Do not declare a second version constant.

1. Bump SCHEMA_VERSION and add the step in src-tauri/src/db/migrations.rs. Pass the &Connection you were given into an _inner helper. Do not call read_conn() or write_conn() from code that already holds a connection.
2. Do not do filesystem, git, or process work while that connection or the writer mutex is held. Prepare under the lock, do the I/O after releasing it, then write back under the lock.
3. New columns need a default or must be nullable, so existing rows still read. Match the steps already in src-tauri/src/db/migrations.rs. Do not add a second migrator.
4. Add a test in src-tauri/src/db/migration_tests.rs that runs the production initialiser on a fresh in-memory database and on a legacy schema that already has rows. Assert the preserved rows, the new columns or indexes, and that a second initialisation changes nothing. A test that uses db::get() installs its database with db::test_support::isolated() and holds the guard for the whole body.
5. Update src-tauri/src/db/schema_dump.txt in the same change. init_schema_dump_matches_committed_snapshot in src-tauri/src/db/seam_tests.rs fails when the fresh schema drifts from that file.

From src-tauri/:

cargo test --lib migration_tests

cargo test --lib init_schema_dump_matches_committed_snapshot