Turn shuffle option on in soca -> cice. (#60)

This turns on shuffle option in soca -> cice and changes `seaice edge`
to 40%.
For all background ice concentrations < `seaice edge` when shuffle is
on, the code finds a nearest background point whose total ice
concentration is closest to the soca analysis total ice concentration,
and copies that background ice profile as the new analysis at the
current point. For all background ice concentrations > `seaice edge` ice
concentrations are rescaled, and some ice fields recomputed.

Some results from a first cycle of high res experiment:
Background:

![ice_bg](https://github.com/user-attachments/assets/645f8826-9cad-4987-8fc4-7cf240ee5b15)
soca analysis (no postprocessing; I am cutting off values outside of [0,
1] when plotting):

![ice_an_soca](https://github.com/user-attachments/assets/e52651ea-e30f-4091-bf88-2becb8ee6c4c)
analysis after Soca2Cice on develop (where `shuffle` is off, and `seaice
edge` for the arctic is 0.8) As expected, the analysis looks like
background for ice concentrations < 0.8

![ice_an_soca2cice_noshuffle](https://github.com/user-attachments/assets/2ba397d5-14a9-451f-bbeb-b5bae4e3c20d)
analysis after Soca2Cice in this branch (`shuffle` is on, `seaice edge`
is set to 0.4). Note that these results are with
JCSDA-internal/soca#1108

![ice_an_soca2cice_shuffle_bugfix_edge40](https://github.com/user-attachments/assets/4b45072c-1ea3-47f6-a488-62f7720bd1fa)

This branch can be merged in develop regardless of
JCSDA-internal/soca#1108, since gdasapp soca
hash already includes the bugfix that allows `shuffle` to be turned on.
However, I would recommend running experiments that include
JCSDA-internal/soca#1108 which further improves
on shuffle.

I ran C384mx025_3DVarAOWCDA for 2 cycles with this setup, and forecasts
succeeded (cice was OK with these analyses). We'll do better evaluation
when we run the next marine candidate experiment.

Co-authored-by: Anna Shlyaeva <anna.v.shlyaeva@noaa.gov>

algorithm/marine/soca_2cice_global.yaml.j2

@@ -9,15 +9,15 @@ output geometry:
 variable change:
   variable change name: Soca2Cice
   arctic:
-    seaice edge: 0.8
-    shuffle: false
+    seaice edge: 0.4
+    shuffle: true
     rescale prior:
       rescale: true
       min hice: 0.5
       min hsno: 0.1
   antarctic:
-    seaice edge: 0.9
-    shuffle: false  # seg. fault when true
+    seaice edge: 0.4
+    shuffle: true
     rescale prior:
       rescale: true
       min hice: 0.5