fix: Address snapshot and sanitization issues in packages/core

[Winify]

What & why

Fixes two classes of bugs in the mobile element snapshot and XML parsing pipelines discovered during focused review for platform gaps (iOS, Android, WebView).

1. Element snapshot: child-text name enrichment (element-snapshot.ts)

The previous code promoted class-based Android locators to text-based selectors by reading child text. This had three defects:

• Wrong element target: UiSelector().text("...") resolves to the child TextView, not the clickable parent container the locator was generated for.
• No uniqueness check: The full locator pipeline validates text uniqueness via XPath before emitting selectors. The promotion skipped this.
• DFS preference for deeply nested text: A nested grandchild label was preferred over a direct sibling label.

Replaced with name enrichment: when an interactive Android element has only a class-based locator and no name, child text is pulled (BFS, length-guarded <100) and used as the element display name. The selector stays class-based, correctly targeting the parent.

2. XML parsing: sanitize page source before parsing (xml-parsing.ts)

Closes #240. iOS page source from XCUITest (especially WebView content) can contain HTML artifacts that @xmldom/xmldom strict XML parser rejects:

• Bare ampersands (& in URLs,  , ©) not defined in XML
• HTML void elements (<link>, <meta>, <br>, <img>) without explicit /> self-closure

Added sanitizePageSource() called from both xmlToJSON and xmlToDOM. Two-pass: escape bare & → & (preserving valid XML entities), then self-close 13 known HTML void elements. Platform-agnostic — runs for both Android and iOS.

3. escapeText: added \r and \t escaping (locator-generation.ts)

Control characters in element text would produce invalid Java string literals in UiSelector().text("...") selectors.

Type of change

• Bugfix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Polish (an improvement to an existing feature)
• Breaking change (existing behavior changes for users)
• Documentation
• Internal (build, CI, dependencies, tooling)

Packages touched

• shared (types and contracts)
• core (framework-agnostic capture/reporting)
• service (WebdriverIO adapter)
• nightwatch-devtools (Nightwatch adapter)
• selenium-devtools (Selenium adapter)
• backend (server)
• app (UI)
• script (page-injected runtime)

Notes for reviewers

• Child-text name enrichment is Android-only for now. getFirstChildText reads attributes.text (Android attribute). An iOS equivalent would need to check name/label/value. Filed as follow-up.
• _sourceXML auto-detection added to jsonElementToSnapshotNodes to match serializeMobileSnapshot behavior.
• Pre-existing bugs found but not fixed: getBestAndroidLocator uses ~ prefix for attrs.text (~ maps to content-desc, not text on Android). Same pattern in getBestIOSLocator for label/value. Fixing would change behavior for existing consumers — needs a separate change.