Trawling or Visual Censuses? Methodological Bias in the Assessment of Fish Populations in Seagrass Beds