Feat/app UI i18n insights contracts

[berrybluecode]

Changes

This PR adds substantially richer
authoring UI, and overhauls the deployment story. At a high level:

Data-contract & AI rule generation

• Import ODCS v3.x data contracts and generate DQX rules (predefined,
  property-based, and natural-language) via a new contract_rules_service and
  /v1/contract route, including a dedicated "from contract" UI.
• Natural-language (type: text) expectations are generated through the app's
  ChatDatabricks Foundation Model path (no dspy/Spark needed in-container).
• Reference-table checks (foreign_key, has_valid_schema) are now authored in
  the single-table editor with a table picker; schema-validation gets its own
  editor and mutual-exclusion handling for expected_schema vs ref_table.

UI features

• Internationalization (English, Spanish, Italian, Portuguese-BR) with a
  language selector and locale-parity test.
• Insights dashboard is now hosted persistently so the embedded Lakeview iframe
  no longer reloads on every navigation; expandable labels UI; component
  reorganisation (apx/ → layout/); regenerated API client (orval).

Deployment & docs

• Updated DEPLOYMENT.md / DEVELOPMENT.md (CLI v1.4.0+ gate, one-button
  postgres_roles, bind workflow) and removed duplicated content.

Linked issues

Resolves #..

Tests

Added backend unit suites for the Lakebase executor, Postgres/Delta migration
runners, contract rule generation, discovery DDL, dry-run dispatch, app
scheduler lease, quarantine filtering, lint policy, and i18n locale parity.
make app-test (658 passing) and make test (1160 passing) are green, along
with make fmt and make app-check.

• manually tested
• added unit tests
• added integration tests
• added end-to-end tests
• added performance tests

Documentation and Demos

• added/updated demos
• added/updated docs
• added/updated agent skills

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.43%. Comparing base (9980c98) to head (cddd404).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1262      +/-   ##
==========================================
- Coverage   92.56%   92.43%   -0.13%     
==========================================
  Files         102      102              
  Lines       10075    10075              
==========================================
- Hits         9326     9313      -13     
- Misses        749      762      +13     

Flag	Coverage Δ
anomaly	54.37% <ø> (ø)
anomaly-serverless	54.38% <ø> (ø)
integration	50.54% <ø> (-0.01%)	⬇️
integration-serverless	50.40% <ø> (-0.23%)	⬇️
unit	57.02% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[github-actions]

✅ 772/772 passed, 1 flaky, 49 skipped, 5h59m42s total

Flaky tests:

• 🤪 test_quality_checker (57.37s)

_{Running from acceptance #4942}

[github-actions]

✅ 194/194 passed, 2 skipped, 6h49m35s total

_{Running from anomaly #1056}

[mwojtyczka]

Additional findings — pre-existing (not introduced by this PR)

While reviewing the backend changes I hit three issues in code this PR touches/neighbours. git blame puts all three at or before this branch's merge-base, so they're pre-existing, not regressions from this PR — flagging here for visibility; fine to fix in a follow-up rather than blocking this PR. (They're posted at PR level because they sit on unchanged lines, which GitHub won't let me anchor an inline comment to.)

1. Scheduler re-fires a schedule every tick if _trigger_run raises — scheduler_service.py:294-302.
next_run_at is only advanced to a future time at the _upsert_tracker call on line 301, which runs after _trigger_run returns. Per-table failures are caught inside _trigger_run and returned as errors, but _resolve_scope / _load_custom_metrics / DB access sit outside that inner try. If any raises, it propagates to the except at line 302 (log-only); since line 275 already wrote next_run_at = now - 1s, the schedule is "due" again on the next tick → a deterministic failure becomes a tight retry loop that keeps submitting jobs. Relevant now because this PR adds new code paths inside _trigger_run. Fix: advance next_run_at (or persist a failed status with backoff) in a finally/except.

2. Monthly day_of_month silently capped at 28 — scheduler_service.py:1089.
candidate = after.replace(day=min(dom, 28), ...) — a schedule configured for the 29th/30th/31st always fires on the 28th while the UI shows the configured day. Clamp to the actual last day of the target month instead, e.g. calendar.monthrange(year, month)[1].

3. Non-TERMINATED terminal states written verbatim as run status — dryrun.py:533-539.
When status.state is INTERNAL_ERROR or SKIPPED, update_run_status(..., status=status.state, ...) writes those raw strings into the run-history status column, outside the normal SUCCESS/FAILED/CANCELED/RUNNING vocabulary (the synthesized path at 505-513 correctly maps terminal → SUCCESS/FAILED). The row does update (not stuck RUNNING), but downstream UI/queries may not recognize these values. Consider mapping INTERNAL_ERROR/SKIPPED → FAILED.

[mwojtyczka]

Mostly looks good, left some comments. There are some pre-existing issues as well.