When working in high bitdepth, manually changing precision to 8bit or 16bit is
an extraenous step, this adds a conversion step - but doesn't permit dithering
- and also doesn't permit using the palette modes of PNG. However it makes
workflows with single precision linear floating point much more usable.