CF spec: need specification of obs and time dimension for 2d datasets

[gauteh]

When lon and lat is a variable of obs or something else, then it is not always obvious which variable gives the times for obs. currently time is determined as having obs as a dimension and having time in its name.

the same is true for obs if lat, lon has even more dimensions (e.g. trajectory), how do we know (except based on names) what the obs dimension is. this is especially confusing when the dataset is mixed between trajectories and other data (e.g. our trajectory + imu_obs data).

One way is to solve this in a similar way as other CF standards with grid_mapping attributes on the variables. This issue should be discussed with the CF-standard.