Test sharding now recovers or fails on targets that enumerate as empty due to flaky runner boots

When running tests with suite-level sharding, the CLI no longer silently drops test targets that appear empty because xcodebuild -enumerate-tests failed to boot their test runner under load. Instead, it tries to recover each affected target by enumerating it in isolation, and fails loudly with a clear error if the target cannot be recovered. This prevents large test runs from silently executing only a subset of their planned tests. Top-level Swift Testing @Test functions (without an enclosing @Suite) are now also recognized during enumeration, so targets containing only free functions are no longer seen as empty.