Short description of the diagnostic
Add a short description of the diagnostic that you would like to add. There a few tricky points.

1. There is a multi-year averaging. There is a similar one which I'd need to use also in the sea ice diagnostics, so I was wondering if there is a sense to add to preprocessor?
2. There is a GEV fitting for all the grid cells to determine probability indices. It is a super slow process in the current implementation. And by slow I mean really slow, it took like 10 days to run this diagnostic back in the day. I had a few thoughts how to improve it:
   • Somehow parallelize the loop. I don't know what is the best way forward: iris.analysis.maths.apply_ufunc? Some parallelization package? Both? Any feedback here is highly appreciated.
   • The GEV are fitted at the native grid and then the probability indices are regridded to the HadEX3 grid. The problem is that pre-processed data is on the whole globe, and the observations are only on continents and not even everywhere on the continents. Thus a lot of ocean cells are just fitted for nothing. I thought of masking out the oceans, but then the half-ones that got regridded will be not used. I asked for an advice, and got a suggestion of getting the weights from ESMPy regridder and masked based on them. Don't know if it's too complex, but I would really appreciate any ideas!

Branch and pull request
For now original diagnostic code resides here

The recipe here and the corresponding README here

Tagging @Row-Bean and @k-a-webb because we might need this diagnostic for our CCCma applications too.