Text this: Shifting the retinal foundation models paradigm from slices to volumes for optical coherence tomography.