Including guidance on using observational uncertainty in CMIP REF diagnostics

## The motivation

Extensive engagement with the observational dataset community has raised some important considerations around whether and how to use observational data uncertainty in CMIP REF diagnostics.

## The proposed solution
At this time, it seems like there are a few things to discuss when a new diagnostic is added to the REF:

1. When a new diagnostic is submitted there is some kind of check to ensure that guidance is provided on using uncertainty information associated with observations.
2. How this guidance is to be provided - a paragraph in the documentation that says one of (a) no obs used for this diagnostic (b)obs data used in the diagnostic has no uncertainty information (c) links or write up on how to use existing obs uncertainties
3. Any other things to consider?

## Additional context

CCI data and possibly other data sets have technical notes packaged with the data and it would be useful to have pointers to this in the diagnostic documentation as well.

@nocollier - Happy to get your input based on ILAMB diagnostics and if we can have atleast one example for this with the REF.