LoginView, SplashView and ServerSelectionView each picked a random
full-screen background on load, so the visual-regression baselines could
never match — every run compared against a different photo and reported a
spurious 82-94% regression. Centralize the selection in ViewBackground and
let tests disable it, so captured screenshots and baselines are stable.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
