Parse targeting context source files to extract field names

[data-sync-user]

Write a parser class that extracts valid targeting context field names from each platform's stored source file, with version awareness.

Following the pattern of experimenter/outcomes/init.py (lazy-loading class with caching), create a class in experimenter/targeting/ that:

• Reads each platform's targeting context file from the local manifests path (configured in settings.py, pointing into features/manifests//)
• Extracts field names using regex or simple string parsing:
• Desktop (JS): Keys from the ATTRIBUTE_TRANSFORMS object (camelCase: activeExperiments, addressesSaved, etc.)
• Android (Kotlin): Field names from RecordedNimbusContext (snake_case: app_version, is_first_run, etc.)
• iOS (Swift): Dictionary keys from the context dictionary (snake_case: is_first_run, region, etc.)
• Supports versioned lookup: TargetingContextFields.for_application(app, version=None) — defaults to latest (unversioned) if no version specified
• Caches results like Outcomes._outcomes

Acceptance Criteria:

• A parser class exists that extracts field names for each of the three platforms
• Field lists are accessible per application, optionally per version
• Parsers are tested with the current upstream file contents
• Parsers handle minor formatting changes (whitespace, comments) gracefully

┆Issue is synchronized with this Jira Task