fix(FR-1713): fix broken E2E tests and improve session test reliability (#4680)

Resolves [FR-1713](https://lablup.atlassian.net/browse/FR-1713)

## Summary

This PR fixes broken E2E tests and improves session test reliability by addressing selector issues, improving test utilities, and adding comprehensive testing guidelines.

## Changes

### Session Test Improvements (`e2e/session.test.ts`)

- **Fixed selectors**: Updated from `vaadin-grid-cell-content` to modern table row selectors
- **Improved reliability**: Better handling of app dialogs and loading states with proper waits
- **Unique session names**: Added timestamp and random string to prevent conflicts between test runs
- **Menu navigation**: Fixed menu item selector from 'sessions' to 'Sessions' (case-sensitive)
- **Better assertions**: Using `table rows` instead of vaadin-grid for more reliable session verification

### Test Utility Updates (`e2e/utils/test-util.ts`)

- **Rewrote** **`deleteSession`** **function**: Complete refactor with modern selectors and better error handling
- **Added search functionality**: Proper session search before deletion
- **Navigation handling**: Ensures navigation to sessions page before attempting deletion
- **Improved cleanup verification**: Better waiting for session removal confirmation

### Playwright Configuration (`playwright.config.ts`)

- **Visual regression exclusion**: Added `testIgnore: "**/visual_regression/**"` to skip visual tests by default
- **Improved reporters**: Different reporters for CI (html + github) vs local (html + list) environments
- **Better CI integration**: HTML report configured with `open: "never"` for CI runs

### E2E Testing Guidelines (`.github/instructions/e2e.instructions.md`)

- **Comprehensive 749-line guide** for AI assistants writing E2E tests
- **Playwright auto-waiting principles**: Explains when to trust built-in waits vs explicit waits
- **Page Object Pattern**: Examples and best practices for encapsulating page logic
- **Selector strategies**: Priority order from semantic to CSS selectors
- **Common patterns**: Login, modal, table, dropdown interaction examples
- **Anti-patterns**: What to avoid (waitForTimeout, fragile selectors, etc.)
- **Complete test examples**: Real-world test structure and debugging tips

## Test Plan

Run the session creation tests:

```bash
pnpm exec playwright test e2e/session.test.ts
```

Verify the tests:

1. ✅ Interactive session creation on Start page
2. ✅ Batch session creation on Start page
3. ✅ Interactive session creation on Sessions page
4. ✅ Batch session creation on Sessions page

All tests should pass with:

- Proper session name handling (unique names)
- Reliable selectors (no flaky timeouts)
- Proper cleanup (sessions verified to be created)

## Impact

- **Fixed 4 broken session creation tests**
- **Improved test execution reliability** by removing fragile selectors
- **Better test maintainability** with updated utilities and guidelines
- **Faster CI runs** by excluding visual regression tests by default
- **Better developer experience** with comprehensive E2E testing documentation

**Checklist:**

- [x] Test reliability improved (no flaky selectors)
- [x] Proper Playwright patterns (auto-waiting, semantic selectors)
- [x] Updated test utilities for better maintainability
- [x] Added comprehensive documentation for future test writing
- [x] Visual regression tests excluded from default runs

[FR-1713]: https://lablup.atlassian.net/browse/FR-1713?atlOrigin=eyJpIjoiNWRkNTljNzYxNjVmNDY3MDlhMDU5Y2ZhYzA5YTRkZjUiLCJwIjoiZ2l0aHViLWNvbS1KU1cifQ

Author: yomybaby