Scale depth by 65535/max_depth for writing to 16-bit png

I hope I've done this right. I also see a min_depth. perhaps it isn't a commonly used feature. We might want to include it too.

app.py

@@ -202,7 +202,7 @@ with gr.Blocks(css=css) as demo:
         pf_prediction_colored = colorize_depth(pf_prediction, cmap=color_map)
         coarse_predictioncolored = colorize_depth(coarse_prediction, cmap=color_map)
         
-        raw_depth = Image.fromarray((pf_prediction*256).astype('uint16'))
+        raw_depth = Image.fromarray((pf_prediction*(65535.0 / max_depth)).astype('uint16'))
         tmp = tempfile.NamedTemporaryFile(suffix='.png', delete=False)
         raw_depth.save(tmp.name)