Check mixed layer in bias correction

[erwinlambert]

A mixed layer with too little stratification can produce odd results in response to, e.g., thermocline shoaling. Whether or not such regions with deep unstratified mixed layers exist in model output and/or observations is the question. If so, possibly the mixed layer should be treated differently, similar to Naughten et al who applied the 'classical' bias correction (similar to ISMIP6) method there.

To be figured out.