Use units_metadata where appropriate.

[aulemahal]

Addressing a Problem?

It can sometimes be ambiguous whether a temperature represents a difference or an absolute measure.

Potential Solution

The CF conventions have a special way to distinguish between the two : http://cfconventions.org/Data/cf-conventions/cf-conventions-1.11/cf-conventions.html#temperature-units

Additional context

This would be complementary to #1804. That PR removed the ambiguous case that resulted in errors with pint. However, the new attribute could help in making the distinction more explicit. See discussion on the PR.

Contribution

• I would be willing/able to open a Pull Request to contribute this feature.

Code of Conduct

• I agree to follow this project's Code of Conduct

[huard]

One issue is that pint2cfunits returns a string, while here we'd need to return a dict with units and unis_metadata.

One option would be to create pint2cfattrs, which would return a dict. This would not break anything.