[CF-metadata] domain for calendar, add_offset, scale_factor

[Benno Blumenthal]

Jonathan Gregory wrote:
>> 1) I was not aware that it was illegal to scale a coordinate variable 
>> using add_offset and scale_factor.
>>     
>
> It has not been envisaged that there was a need for it, as far as I remember.
> Coord vars usually take a small amount of space, relative to data vars, so
> don't need to be compressed. In principle I don't see why it shouldn't be
> allowed, as they would still satisfy requirements of being monotonic and not
> having missing data. Do you have a need for it?
>   
On occasion we use it for time for time series with many points:   
putting a large offset into add_offset so that time can be single 
precision and still be meaningful.   Obviously double-precision time is 
an alternative, as is changing the offset in the units.   Bottom line is 
that we (IRI aka Ingrid) have always supported add_offset and 
scale_factor for both Coordinate and Non-Coordinate variables, and it 
had not occurred to me that we were doing something unusual.

Benno