Resolve whether slight CMIP differences are due to round vs trunc

Glance review of CMIP output is very, very close, but the question arises whether there's a different treatment of scale and offset conversions between cmi_changer and PDA.