alembic: make migration chain SQLite-compatible (fresh upgrade)

Five existing migrations used op.alter_column / op.create_unique_constraint /
op.drop_constraint / op.create_foreign_key directly on the users + quotes +
quotes_daily tables. SQLite has no native support for those operations and
requires Alembic's batch_alter_table copy-and-rename workaround.

This wasn't noticed until now because the test suite uses
Base.metadata.create_all to materialise schema, not the migration chain
itself; and prod is MariaDB. But running `alembic upgrade head` against
a fresh SQLite database (developer onboarding, CI smoke tests, the
test container's own bootstrap) would fail at 0005.

Fixes:
- alembic/env.py: set render_as_batch=True when the dialect is SQLite.
  This auto-wraps any future autogenerated migration but doesn't
  retroactively rewrite existing op.* calls.
- 0005 (widen quotes.symbol), 0013 (referrals), 0018 (polar webhook),
  0019 (stripe), 0023 (users.lang index + qd_symbol widen) explicitly
  wrap their problematic ops in `with op.batch_alter_table(...) as bop`.

Now `alembic upgrade head` + `alembic downgrade base` round-trip cleanly
on a fresh SQLite database. MariaDB prod behaviour unchanged.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

alembic/env.py

@@ -44,10 +44,17 @@ def run_migrations_offline() -> None:
 
 
 def do_run_migrations(connection: Connection) -> None:
+    # render_as_batch is required for SQLite, which doesn't support
+    # most ALTER COLUMN / ADD CONSTRAINT operations natively. With
+    # batch mode enabled, Alembic emits a copy-and-rename dance under
+    # SQLite while still producing plain ALTER on MariaDB / Postgres,
+    # so prod migrations are unchanged. Detect via the dialect name.
+    render_as_batch = connection.dialect.name == "sqlite"
     context.configure(
         connection=connection,
         target_metadata=target_metadata,
         compare_type=True,
+        render_as_batch=render_as_batch,
     )
     with context.begin_transaction():
         context.run_migrations()