Exact inference on gapped and windowed stationary Gaussian data — gap-pattern generation, windowed frequency-domain covariances, and a hierarchy of likelihoods from Whittle to the exact time-domain solver, matrix-free at scale via preconditioned conjugate gradients
