Merging 'coordinates' can be slow, could be skipped? (performance)

[johnomotani]

In some routines, for example BoutDataset.interpolate_parallel(), many variables have to be merged back together. If they have coordinates, then these are checked for consistency. If coordinates are removed from the variable(s) being merged, then the merge can be much quicker. In some cases (e.g. looking at a slice with only one time point) this merging seems to take longer than the actual parallel interpolation. It might be worth (to improve performance) removing coordinates before merging, or using one of the xarray options ("override" somewhere?) to skip the comparisons - possibly controlled by a global option?

[johnomotani]

Would be interesting to see if this is improved by the latest xarray release - pydata/xarray#7824 may help things.