feat: BDD test suite with playwright-bdd for UX regression testing

[NP-compete]

Summary

• Adds 36 Gherkin feature files with 162 scenarios (~394 generated Playwright tests) using playwright-bdd
• Covers all 25 Studio UI pages, all 6 user roles, all major API endpoints, and visual regression baselines
• Existing 91 Playwright tests are untouched — coexist via dual Playwright project config

What's covered

Domain	Feature Files	Scenarios
Auth & RBAC	4	Login (6 roles), nav visibility, API access control, API keys
Agents	6	Chat creation, deep agents, detail view, agent chat, deploy modal, templates
Memory	3	Short-term CRUD, long-term embeddings + search, compaction
Guardrails	2	Pipeline checks, rule management
Studio UI	15	Dashboard, Connectors, MCP, Models, Skills, Evaluations, Flow Builder, Marketplace, FinOps, Audit Explorer, Approvals, Build Hub, Settings (6 tabs), YAML Editor, Org Chart
Enterprise	3	Multi-tenancy, 5 compliance frameworks (SOC2/GDPR/HIPAA/ISO27001/EU AI Act), audit integrity
Cross-cutting	3	Command palette, theme toggle, visual regression baselines

Architecture

e2e/
├── features/          # 36 Gherkin .feature files (readable by product/QA)
├── steps/common/      # 6 reusable step definitions (auth, nav, api, form, table, visual)
├── steps/domain/      # 5 domain step definitions (agent, memory, guardrails, enterprise, studio)
├── fixtures/          # playwright-bdd base fixtures + page objects
├── tests/             # Existing Playwright tests (UNCHANGED)
└── playwright.config.ts  # Dual project: "bdd" + "existing"

How to run

cd e2e && npm install

npm test              # BDD tests only
npm run test:existing # Original tests only
npm run test:all      # Both together
npm run test:smoke    # Smoke subset (@smoke tagged)
npm run test:visual   # Create/update visual baselines

Test plan

• npm install succeeds in e2e/
• npx bddgen generates .features-gen/ without errors
• npm run test:existing — existing 91 tests still pass
• npm test — BDD scenarios run against running platform
• npm run test:smoke — smoke subset runs independently
• Visual baselines created on first run with --update-snapshots