Implement validator integration in Config Service save flow

Model the return type using a sealed class pattern: `sealed class ConfigSaveResult {}` with subclasses `ConfigSaveSuccess(IntegrationConfig config)` and `ConfigSaveFailure(List errors)`. This enables exhaustive switch matching in BLoC event handlers. The validator call should be `final validationResult = await validator.validate(config);` followed by an early return if errors are present. Never use try/catch to absorb validation errors — they are expected outcomes, not exceptions.

If the validator itself throws (e.g., network error fetching adapter capabilities), catch it and surface as a `ConfigSaveFailure` with a single error of errorCode='VALIDATION_UNAVAILABLE' rather than propagating the exception. For the BLoC layer, ConfigSaveFailure should map to a state that populates field-level error hints on the setup wizard form — ensure the ConfigValidationError.fieldPath values match the form field identifiers used in the wizard UI.

Unit tests covering the save flow integration: (1) valid config passes validation and is persisted — Supabase client insert/upsert is called exactly once, (2) config failing credential validation returns ValidationFailure with errorCode='CREDENTIAL_INCOMPLETE' and Supabase client is never called, (3) config failing field mapping returns ValidationFailure with errorCode='MAPPING_INCOMPLETE' listing missing fields, (4) config failing sync schedule returns ValidationFailure with errorCode='INVALID_SCHEDULE', (5) multiple validation errors are all returned together (not fail-fast), (6) validator throwing unexpectedly is wrapped and re-surfaced as a domain error not a raw exception, (7) update flow applies same validation as save flow. Use a mock IntegrationConfigValidator that returns configurable results. Verify the Supabase mock's insert was called 0 times on any validation failure.