fix(db): make deploy's dry-run clone schema-only by default

[nipunappri]

Branch: fix/deploy-schema-only-clone → main

Summary

• postkit db deploy cloned the entire target database, including every row, into a throwaway local container before running its DDL dry-run — then dropped it. On any database with real data this dominated the command's runtime.
• The dry-run verifies that committed DDL applies cleanly, which only needs the target's structure. The clone is now pg_dump --schema-only by default.
• Added --with-data to opt back in for migrations whose risk lives in the data rather than the schema.

Changes

• db deploy clones structure only by default. deploy.ts now passes schemaOnly: !options.withData to both clone paths (host and docker exec). Previously it passed nothing and inherited the = false default, while db start correctly passed true.
• Added --with-data to db deploy (modules/db/index.ts) for cases where the migration's risk is in the data: adding NOT NULL or UNIQUE to a populated column, type narrowing, backfills.
• Made schemaOnly a required parameter on cloneDatabase and cloneDatabaseViaContainer. The silent default is exactly how this bug happened — one caller remembered, one didn't. Requiring it means no future caller can pull row data by omission. Typechecking confirmed start and deploy were the only production callers.
• Warn on --with-data that seeds run as part of the dry-run against a clone that already holds the target's rows, so seed inserts may conflict on existing keys.
• Clone progress messages now state which mode is running ("no row data" / "with row data") so the cost is visible rather than inferred from how long it hangs.
• Docs (cli/docs/db.md, CLAUDE.md) updated — the old behavior was documented as deliberate, so those passages needed rewriting, not just extending.

Type of Change

• fix: Bug fix

Test Plan

• Unit tests pass (npm run test) — 382 passed, 31 files
• Build succeeds (npm run build)
• Typecheck clean (npx tsc --noEmit)
• Added smoke test asserting db deploy --help documents --with-data as opt-in
• Updated service-level tests that asserted the old implicit default; the explicit schemaOnly: true / false argument paths are both still covered
• E2E tests (npm run test:e2e) — not run locally (needs Docker/testcontainers); please confirm in CI
• Manually verified against a remote with a non-trivial dataset — worth one reviewer pass before merge

Breaking Changes

• Behavior change, not an API break. postkit db deploy no longer clones the target's row data during its dry-run.

cli/docs/db.md previously documented the full-data clone as intentional — "it still clones full data, since it's specifically verifying real migrations against realistic data before they touch production." That rationale is real, so the capability is preserved behind --with-data rather than removed.

The trade-off: a migration that would only fail against real rows (e.g. ALTER COLUMN ... SET NOT NULL on a column containing nulls) will now pass the dry-run and fail on the target instead. Anyone relying on the old behavior should add --with-data to their deploy invocation. Worth calling out in release notes.

The public signatures of cloneDatabase / cloneDatabaseViaContainer gained a required parameter. Both are internal to the CLI, so this only affects in-repo callers.