Why moving embedding inside turbopuffer drops search latency

Native embedding isn't just a convenience feature, but a way to shave hundreds of milliseconds off your search queries.