major variables add (GradientFeatures)

[vincelhx]

• streaks.py
  => added heterogeneity mask (owiWindFilter) + 400VV & 800VH masks in xr_dataset
  (Yuan Zhao 2021)

=> added work on streaks from gradients (Koch 2004) with mulitple polarisation/windows_size/resolutions results

• inversion.py
  => started a work to have a logger
  => deleted some useless .compute() ?! maybe less time for computation
  => added owiAncillaryNrcs and owiAncillaryNrcs_cross to simulate NRCS from ancillary wind using gmf

• config
  => small changes in the config file :

add_gradientsfeatures: True
add_nrcs_model: True

if not specified both are set to false

• results saving
  owiWindFilter (mask) + 4 temporar variables are added in the main group
  a group is created for directions results : owiWindStreaks

• I also had to change RCM output format

[agrouaze]

What is the rational of having 2 handlers rather than one?

[vincelhx]

changed

[vincelhx]

also changed in xsarsea to

logging.basicConfig(level=logging.INFO,
                    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
logger = logging.getLogger('xsarsea')

[vincelhx]

we thought that inversion time was too long : sometimes 10 minutes for a EW

I added a xr_dataset.load() at the begining of the use of xr_dataset. It looks to be 5x faster than without it.
It forces the load of data in memory and executes differ operations before inversion and streaks operations